This is an excellent use case for completely local, small model inference, yet for inexplicable reasons Mozilla wants to normalize uploading your entire private browsing history to a cloud.
These (Mistral's and Mozilla's) marketing pages aren't candid enough to clearly explain the difference between local and cloud inference, and that they're asking you to consent to enabling the latter. I'd call that the bare minimum of ethics. If you're in a position of authority over your less technical users, you are ethically obligated to give them the full picture of what are you doing and why they should consent.
(Aside to any Mozilla people who might be reading HN, your page here[0] has an oversight, it advertises this model as "Mistral Small 4" but the hyperlink is to OpenAI's model card for gpt-oss-120b).
I am not trying to defend Mozilla doing this, and I don't support sending data to cloud based services like this in a way that users won't understand.
But I also think that the state of the art in small LLM and user device capabilities aren't there yet to put a "good enough to be actually useful" local-only LLM as a prepackaged thing in a mass market distributed browser.
You don't want a browser that takes 10GB of extra RAM (on top of the memory hog that is having just 3 or 4 complex tabs open on its own already) and pegs your CPU at 99% usage for minutes at a time. And not in an era when mass market consumer laptops are still commonly 8GB or 16GB of total system RAM. Many of those with integrated-into-CPU onboard graphics (eg: not a gaming laptop with a discrete GPU on the PCI-E bus).
It'll be a catastrophe for laptop battery life, among other resource use problems. And an LLM that fits in under 8 to 10GB of RAM for CPU-only inference is not going to be nearly as capable as an off-device inference system.
I wish they had just done this with a very clear up front opt in (not enabled by default) thing that explains what Mistral is, that it's not some big American cloud company but a relatively small startup in France, and that your prompts/LLM interactions will go to their servers. And some documentation on how it will be handled/stored in a supposedly trustworthy manner.
Well, it depends on the task, doesn't it? "running shoes I looked at last week" / "Here's what I found in your browsing history:" doesn't need a 119 billion parameter frontier model; it's a RAG problem for the 0.6 B embedding models. That's an example Mozilla offers. Presumably to explain to their users why it's essential they hand over their last week's browsing history for this convenience (but it isn't! Hardly for that!)
I feel it's wrong to tell users that it's important and normal to relinquish all control of their—extremely personal—life history, in bulk, in plaintext, to strangers.
I agree wholeheartedly that remote server inference is super useful, and that local inference falls far short on many tasks. (I have no objection at all to Mozilla providing a cloud inference feature).
What I don't buy is that we must ask users to redraw their personal boundaries so that their most intimate life details, and remote frontier-model inference, overlap. They do not need to overlap.
You can accomplish a lot with private local inference with the smallest of models; and you can accomplish a lot on remote servers which aren't privy to everything. If some convenience is lost by not combining the two, well, so be it. I'm sure most people would agree, if all of this was laid out plainly.
You’re taking their shoe example way too literally. It’s just showing they’re catering to the average user, not an engineer. And tons of average Joe users now have very high expectations of how intelligent a LLM is because they’ve interacted with ChatGPT and the like. If there’s a super tiny model that is too dumb to do anything beyond shoe searching and they have to switch to Gemini and google AI search for anything more complex, this whole launch would be an immediate failure.
I'm of the opinion that local inference should be done to the greatest extent that is realistically possible, at the earlier time that the hardware/average user platform is capable of doing so. I personally spend a fair bit on kWh extra in my home electrical bill monthly for having a good sized chunk of local inference ability in my house, but that's not a common thing yet.
If mozilla is doing things to send users down the path of doing this externally, they need to be much more upfront and transparent with the users about where their data is going, and not bury it in some terms/conditions that only nerds will hunt for.
Not really, but it's probably easier to make it on top of LLM than to make specially-purposed tool for it, if we talking in terms of time-to-market effort.
SQLite FTS could have done this a decade ago. We've had good local search capabilities for two decades and they either been underused or abandoned (e.g. Google Desktop, Yahoo! Search). This may be a reasonable projection of where AI is headed. You can do a lot locally but there is too much incentive to centralize around cloud infrastructure, then the privacy concerns make that prohibitive and we end up with what feels like a false choice of cloud or bust.
Maybe it's LLM hype that will bring more powerful capabilities to the desktop?
I think classification and clusterisation is more important than search per se. "Where did I see that article about that weird psychological effect where people remember things they haven't seen" is not resolved by direct search, but can be - with some work - helped by language models and NLP. One doesn't need full blown frontier LLM for that, but bare word indexing would probably not do either.
Firefox is already using small local models for a few things, like tab grouping and link previews. I assume that this is for cases where a reasonably sized model isn't adequate.
> Firefox is already using small local models for a few things, like tab grouping and link previews
Which I immediately disabled alongside their "improved" Nova design, cause it doesn't improve productivity. These are all distractions from web browsing. If they want to improve anything, they should work on their abysmal spell checker, but that's not shiny enough.
Let's say the goal is doing something for work, so accuracy is important, and you prefer for it to be a nice reading experience, like a good translation.
As long as cloud model are somewhat better in these things, it's good that users would have the option to use cloud models.
Dont use this for work unless it has been verified by your corporate security and legal teams. Sharing business data with a company with which you dont have a enterprise agreement can land you in serious trouble.
> that it's not some big American cloud company but a relatively small startup in France
How it's any better? Small companies can be bought by big companies. French government is as capable of trampling over their citizen's privacy the moment they feel they need it as US government is. And putting a squeeze on a small startup is way easier than on a major cloud company (not that either is particularly hard). Also, OpenAI used to be an idealistic non-profit one day too, then it started to smell trillions and all that went of of the window.
> is not going to be nearly as capable as an off-device inference system.
I rarely need PhD-level research into my browsing history. I'm not going to solve millennium problems on my bookmarks. The tasks that I will realistically need are well within capacity of most very basic local models. Maybe they'd be a bit slower, who cares.
Let's not pretend that EU data laws are not far more stringent and actually enforced than whatever the US has. Like sure, the government with a jury decision may acquire some person's data. But it's not getting sold to whichever-random-company-offered-the-highest.
Also, some shady TOS is also not enough to override the law in the EU, you have much more protection as a user here.
> But it's not getting sold to whichever-random-company-offered-the-highest.
Nothing in GDPR prevents transfer or selling of personal data. True, there are hoops to be jumped to do that, both in terms of consent and documentation, but let's not pretend a lot of people would read the welcome banner and refuse to interact with a service because its legalese says "we'll sell you data and if you don't want us to, go away". Yes, selling the data becomes more expensive because you need to hire the lawyers to produce those welcome pages and regulatory compliance paperwork, but large corps have enough money for lawyers.
> Also, some shady TOS is also not enough to override the law in the EU
It doesn't need to override anything, as the law does not prohibit data collection or transfer. It only describes the hoops that need to be jumped to achieve it.
It certainly does. You need to weigh the users privacy with your interests and can’t just put them aside. Noncompliance leads to quite large fines. You can easily find examples of fines given.
The DPA’s are short on capacity, sure, and large corps will try to fight the fines in court. But to describe it as “just hoops whilst allowing everything” is unfair.
> You need to weigh the users privacy with your interests and can’t just put them aside.
Who said anything about putting them aside? There would be a process and highly paid lawyers confirming that selling user data is only for the user's ultimate benefit, as it allows to provide awesome services to the users, and it all will be outlined in a 100-page privacy policy which you will be sure to read on each site you use, wouldn't you?
> But to describe it as “just hoops whilst allowing everything” is unfair.
Can you quote me the place where GDPR prohibits this? Not says something like "weigh user privacy" and "take adequate measures" and so on - which can always be resolved as "we weighed carefully and we took measures and we decided selling the data was awesome and users love it" - but explicitly and unambiguously prohibits the practice? If you don't find it - that description is exactly what it is.
>How it's any better? Small companies can be bought by big companies. French government is as capable of trampling over their citizen's privacy the moment they feel they need it as US government is.
It's better for Europeans because the company in question is European and so will not export their data to New Jersey. And because any success it has will presumably better benefit France and Europe.
> It's better for Europeans because the company in question is European and so will not export their data to New Jersey
There aren't many government institutions in New Jersey that would take a lot of interest in anything Frenchmen are doing. There are a lot of institutions in France that would be interested in anything Frenchmen are doing, if it contradicts what the French government wants to be happening. The main threat to citizen's privacy always comes from the government closest to them, the government overseas has its own citizens to worry about and pays much less attention to foreign citizens on foreign land. There could be exceptions, true, but as a rule, if you look into New Jersey, you'd sooner find mass surveillance of American citizens than mass surveillance of the French.
Same goes for commercial interests. If I want to run targeted ads in New Jersey, I want to have profiles of New Jersey people, not French people. So buying data from France in New Jersey would not be a routine occurrence, but buying local data would be.
> How it's any better? Small companies can be bought by big companies.
This. I've come to view a startup as a company without a business model, doing everything to get acquired by a mega corp that will finally squeeze the juice out of the userbase
(The lack of a viable business model applies to some mega corps too)
>But I also think that the state of the art in small LLM and user device capabilities aren't there yet to put a "good enough to be actually useful" local-only LLM
so then don't add it in as highly advertised feature until it is. doing things right and living up to your core values is a lot to expect from businesses these days but, at minimum, a non-profit foundation should be able to live up to these goals, yes?
But apple said their AI only falls back to the private cloud when it has to? Just kidding, near every request needs to fallback, because a phone can't actually run a real LLM, no matter how many "neural cores" it has..
The only people having an expectation of translations being done locally is exactly the nerds that keep whining that it's not using a local model. Every single normal person, when presented with a "translate" button either know it's going online, or don't care about it.
Begging the purists to run away from Firefox at this point so they can stop wasting everyone's time. Your demands for examplarity and whining about money not going ONLY to firefox and jerking yourselves on Servo was not enough, now you want to restrict the browser to owners of an RTX5080 if they want to use it?
You realize that people that don't care won't even know what Firefox is? Being a niche player and scaring away niche people is the most stupid strategy possible.
Being a small player and throwing away any hope of expanding your user base by building for people who will complain no matter what you do is, well, something!
Ah yes, the customers of Mozilla, who are certainly paying users and not just people whining and saying BACK IN MY DAYS IN THE LAST MILLENIUM IT WAS DIFFERENT.
Get over it gramps, things are changing. Mozilla spent ten years trying to mostly appeal to these people by staying very conservative, and it got them nothing but abuse and harassment from people not realising that staying stuck in the past is a death sentence for Mozilla.
You want to save Mozilla, burn down Google first. The ecosystem in which they are evolving isn't driven by them.
I solemnly proclaim I would pay reasonable money for a browser (Mozilla or otherwise) if their maker would truly try to make the best browser possible and succeed at least partially at it, instead of serving a thousand of agendas none of which is making the best browser possible. The sole reason why I am not Mozilla's customer is exactly that - because they do not see me as their potential customer and don't care about my needs. Of course, they are free to do that but the result is people like me aren't giving them their money. So they have to go to Google.
> You want to save Mozilla, burn down Google first.
Uhhh... 85% of mozilla's revenue comes from Google via the search deal. In a theoretical scenario where a genie waved a magic wand tomorrow and google went "poof", disappearing, mozilla would be in a dire emergency for lack of revenue.
It would. But Mozilla has zero long term survival chance in the current ecosystem. The economics are just not there, there cannot be more than one big browser in the way this system works.
I do not fundamentally disagree with you but it's also an extremely well known phenomenon that average non tech users, in the aggregate of millions of people, will click almost any "yes/I agree/continue/Next" step on a software installer or new user sign up workflow for anything, without reading the ToS. People blithly sign up for all sorts of cloud based things and services without understanding their full ramifications all the time.
I also wonder at the specific level of aggression and the tone of your comment which does not seem to be an appropriate response to that person's specific comment.
People such as you are describing and rightfully criticizing are knowingly taking advantage of that. Indeed it's how a lot of malware gets installed too.
The person you're responding to is pointing out that a lot of people at the surface level do only appear to care about the results. They put something into google translate, it works, they gets results they are pleased with, they don't put a lot of thought into the fact that the data is going to an external service. That's not an inaccurate description of how a lot of people use their computers these days. Look at how many people signed up for ChatGPT accounts and put the chatgpt app on their phones and talk to it all the time. That's the level of critical thinking a lot of non tech users have about their personal data.
The fact that people will click yes/agree/OK on almost anything is how Windows computers got Bonzi Buddy installed on them back in the day, and now it's continued into the cloud-everything era.
> it's also an extremely well known phenomenon that average non tech users, in the aggregate of millions of people, will click almost any "yes/I agree/continue/Next" step on a software installer or new user sign up workflow for anything, without reading the ToS. People blithly sign up for all sorts of cloud based things and services without understanding their full ramifications all the time.
I would put a lot of that down to learned helplessness.
A lot of non-tech people tell me that "they already know" and that it is impossible to avoid. I have been told I am naive to think its possible to keep data private.
It's almost as if outside the tech bubbles most people don't understand the tech or implications. HN posters who live and breath in this space are so smug to pretend this esoteric knowledge about networking identifiers or fingerprinting is simple and that anyone can get it. These are the same people who will throw their hands up on any other esoteric topic and give up on it equally. For example, tech people so easily decry how "stupid" the law and legal jurisprudence is (because obviously if they don't understand it, it must be bullshit) but see zero irony in being smug pretentious assholes to regular people who didn't happen to major in computer science.
We shouldn't all have to be PhDs in law and jurisprudence to trust a legal system to look after our interests just like we shouldn't all have to be computer scientists to trust that technical stewards are not exploiting people's lack of knowledge.
Sick of this mentality. Oh these people are sheep idiots who don't understand the EULA euphemisms of "improved user experience " means adtech and behavioural tracking.
and at least one more in another thread today. This is not what HN is for and it destroys what it is for. You may not owe Mozilla better, but you owe HN better if you want to participate here. This is only a place where people want to participate becuase enough people make the effort to raise the standards rather than drag them down. https://news.ycombinator.com/newsguidelines.html
That's a lot of words and assumptions, when a single check of my posting history would show that I despise the HN bros as well. Jumping on a tangent about libertarianism when you could have simply called them retarded and saved a lot of time.
Anyways, no, I'm talking about the average person, the public worker, the person that thinks the internet is the funnily named Safari app, the elderly: they give zero fucks about it going to some service online. They used to search for Google translate before, whether or not it goes on someone's server, they do. not. care. You're not going to win them over with "it runs on your device". Their device is a crappy laptop that barely runs excel, and if they can offload computing, they will.
Have you tried Ling-3.0-tiny? It runs fine on CPU-- on a 14700KF gets 40tg/s and 250pp/s and on a ordinary gpu (RTX 4070) does over 200tg/s with no MTP and 7185pp/s. (my figures are Q8, though presumably a good Q4 would be faster)
It's certainly not as capable as something that needs a high memory gpu for quick performance, but I was quite impressed with it for what it is.
(and fwiw, I had it translate your last paragraph to German, then used google translate back to english: "I wish they had handled this clearly and transparently via an opt-in mechanism—not enabled by default—that explains what Mistral is (not a major American cloud company, but a relatively small French startup) and that your prompts and LLM activities are sent to their servers. I also wish there were documentation explaining how the data is handled and stored in a way that inspires trust.").
How much RAM does it take up in total? I'll have to give that a try on one of my test systems. Looking at a somewhat randomly chose GGUF quantization of it, looks like just under 5GB on disk in Q4, so RAM usage somewhere around 5-6GB?
it's an extremely sparse MOE, so there is some odds of acceptable performance using a smaller in-memory cache and the rest on flash. ... I don't have a setup to test that right now.
(Of course, if translation is all you want much smaller models will work. Ling-tiny can do summarization, dom manipulation, scripting, etc. too).
> You don't want a browser that takes 10GB of extra RAM (on top of the memory hog that is having just 3 or 4 complex tabs open on its own already) and pegs your CPU at 99% usage for minutes at a time.
But I do, especially when the choice is either having it locally, remotely, or not at all. I also indeed do nit my CPU used for such, but GPU. I could even run a 8 GB model on a remote (but still local network, on-prem) NPU.
There's one caveat though: if you are gaming and browsing.
> But I also think that the state of the art in small LLM and user device capabilities aren't there yet to put a "good enough to be actually useful" local-only LLM as a prepackaged thing in a mass market distributed browser.
Then just allow it to be enabled on high end devices? But it must be local only. As hardware advances and people upgrade, more people will be able to turn on the feature.
> I wish they had just done this with a very clear up front opt in
If a local model is not realistic, then this should not have even been an in-your-face opt-in, but at most some add-on.
Of course, their telemetry isn't even opt-out, so even the opt-out for the Mistral thing is kind of disingenuous on their part, since they get a bunch of information from us in other ways.
(sigh) Ah, Mozilla has gone down such a dark path over the years. Too bad.
I'm worried that the middle could fall out of the computing market across the board. If you can afford to keep up with the upgrade treadmill, you'll get private, local inference capabilities. If you can't afford to stay on the treadmill, you'll be stuck with whatever cloudshit malware Silicon Valley wants to foist on you.
I acknowledge that this is already the case, to some extent. The cheapest laptops at Best Buy are crammed with the most preinstalled malware. That's been the case for, what, 25 years? But you've always been able to wipe that cheap laptop and make it into a much more capable, trustworthy machine.
Well, assuming LLMs do become a pervasive part of the computing experience, what happens to the cheap laptops? Do all computers get more expensive to accommodate local inference? Does the rift between the everyday user's experience and the savvy user's experience grow even wider than it already is? Neither outcome seems good for the average joe who just needs to check his email.
Remember how in like 1999/2000 Sun was trying to predict that everyone's computer would be some form of thin terminal in the future? Turns out they were very wrong on the part about it running on Sun server back-end infrastructure, but that same general purpose has now been accomplished through other methods where a lot of people do basically EVERYTHING inside a web browser tab to some external cloud service.
Now add the need for external inference because very few random consumers are going to buy a $3000 laptop when they can get the $600 laptop at Best Buy, and that trend further escalates.
This has always been the case, forever. You have to pay for a product or service. How you do so can be with cash or your data/body/vote/eyeballs/indirect discretionary purchases.
The amount of work that can be done funded by foundations and free work is nowhere close to what people want.
Oh give me a break. This argument that people's objections to advertising comes from some Pollyanna naivety over things being free is nonsense. Tell me the last time you have ever seen a company be upfront and explicitly offer a free tier where they tell you upfront exactly what they are harvesting about you and for what purpose (and no, "improving user experience" isn't being upfront) but also offer you a paid version where they explicitly promise not to do that.
"If the product is free then you are the product" is supposed to be a cautionary observation, not an axiomatic proscription.
Does anyone actually trust companies that offer paid services to not harvest their data? Whenever I see this trope I think to myself that whatever lack of regulation, oversight and enforcement lead to that being okay would equally allow for them to both take my money and harvest all my of data anyways.
This is akin to people who criticise socialised medicine or services and pejoratively characterize supporters as just wanting "free stuff", as if the concept of collective payment is naive or something
The privacy policy for Smart Window[1] indicates that prompts are not only sent to an LLM provider (Mistral in this case), but also to Mozilla servers first.
> The assistant sends the full prompt (including your query, any relevant Memories, and any additional relevant browsing context) to a Mozilla server.
>
> Upon receipt, Mozilla forwards the request to a third-party large language model (LLM) on your behalf. The LLM receives the request from Mozilla, not directly from you, and sees a Mozilla IP address rather than your own. The request includes only the information needed to generate a response or perform the requested action.
Where they do use an on-device model is for the initial query classification:
> When you submit a query in the Smart Bar, the assistant uses a local (on-device) intent classification model to determine whether the query is best addressed by a chat or a search.
> Smart Window infers information about you based on your interactions with the assistant and your Firefox browsing activity, including websites you have visited. This may include activity outside of Smart Window (such as in classic windows)...
Privacy-minded users who want to try this feature would need to create a totally separate browser profile to avoid data leaks.
Some years ago Firefox was the go to browser if you wanted to have "some" privacy in your browsing (together with uBlock origin). With news like this I m really wondering if my views are outdated and I need to to some good researching on maintaining some privacy in what I m browsing.
Mozilla has been in a bit of a recursive feedback loop death spiral for years now of: Browser loses market share -> try some weird thing -> very few people like or use it -> browser loses market share -> management says "oh shit we're losing market share we better try some weird thing"
I say this as a person that uses firefox with ublock origin 99.5% of the time. It's better in my opinion than Chrome. I can at least easily turn off the enabled by default crap features. Like the advertising and sponsored news links on the default new tab page.
This is the confidently incorrect narrative that gets repeated in the comment sections practically every time Mozilla is mentioned but it's every bit as incorrect now as it's been all the previous times.
So here we go again for the millionth time: the big losses of Mozilla market share were approximately during the 2010s. The era of side bets on unique features is approximately the 2020s. The unique features didn't retroactively cause the market share losses of the 2010s.
Moreover, telling the market share story in terms of specific browser features misses the elephant in the room, which is that Google, with the world's most visited page, and a browser that's the installed default on over a billion devices, grew it's market share with a combination of web visability and dominance over the most used mobile platform. Mozilla could triple their budget and have the world's best browser experience, but it wouldn't make much of a dent against distribution defaults.
I would wager that the impact on market share is driven about 97% by Googles distribution advantage and 3% by aligning with user preferences on features and performance. If being a perfect browser led to market dominance, Opera would have already conquered the world back in 2012, but the economics of building a browser aren't always friendly to the good guys.
FF's market share crashed because it was painfully slow compared to chrome when it first came out. The marketing helped, but it wouldn't have stuck if it hadn't been better. You can only blame users being too stupid to know better for just so long.
I never said that Chrome and Edge don't have a huge distribution advantage due to being installed by default on peoples' devices. That's also certainly a huge factor in the mass adoption of Chrome as what people consider "the web browser. Or Safari as default browser in MacOS of course.
Because just doing a browser is not working. Market share keeps declining. And Firefox is pretty good. They spend the majority of their resources on Firefox, and always have.
I very much think it did, because FF once was recommended by us nerds, we installed it for people and praised it. And that made an impact outside of our circles.
Now I still occasionaly install it along with ublock origin because there is no alternative, but I don't praise it anymore (or bother to install it for someone in the first place) - but rather bitch about how they also sneak in advertisement and spyware.
I think it didn't help. Constantly adding or changing features makes it less consistent and predictable. For example several family members were confused by what is "pocket".
They have twice (that I'm aware of) tried rebranding as an advertising company. They are funded by Google (though not _necessarily_ influenced by them). There was an article in LWN by them about (in part) how they can't get enough information about users of Thunderbird so they were suggesting telemetry should be enabled by default and opt out instead of opt in because most people won't opt out.
I think they get a lot of credit because they aren't Google and not enough push back
It's more un-fathomable if you recall Mozilla were the people who *pioneered* local browser ML as a privacy feature. Back in 2022—when Chrome was doing language translation by uploading to Google servers, and Mozilla introduced the (actually innovative!) idea of baking small ML models into the browser,
And look: they were 100% cognizant, then, of how vital local processing is to privacy:
> "...called Project Bergamot. The ultimate goal of this consortium was to build a set of neural machine translation tools that would enable Mozilla to develop a website translation add-on that operates locally, i.e. the engines, language models and in-page translation algorithms would need to reside and be executed entirely in the user’s computer, so none of the data would be sent to the cloud, making it entirely private."
Even more un-fathomable, Mozilla's CTO has been writing a substack with the underlying theme of "if the AI isn't running on your hardware, you don't own it" https://newsletter.ownersnotrenters.com/
You cannot ship on device inference with a real LLM still. Average people freak the hell out if it's even 10% slower than whatever google ships. At least thats the claim for how Firefox lost all their marketshare in the first place (Which was not the case. When everyone was claiming firefox was "slow", it simply was not, as long as you used an ad blocker. Normal people like my dad didn't switch to Chrome. It was installed through a sketchy mechanism and they never noticed)
Meanwhile, if you CAN run these models locally, using your own hardware for inference is supported. It's a setting option.
Google still blasts you to install Chrome at every opportunity if you use their services, and is heavily optimized for Chrome (e.g. Docs, YouTube). That's not really Firefox's fault.
> This is an excellent use case for completely local, small model inference, yet for inexplicable reasons Mozilla wants to normalize uploading your entire private browsing history to a cloud.
I certainly agree that this option would be nice to have, but running firefox without an ad blocker on the modern web will ALREADY spin up my fans just rendering all the garbage, and I have a decent laptop.
You can't expect the median Firefox-running device to be able to run a capable enough local model effectively.
Indeed. Realistically a 'capable' small local LLM, even one that's definitely not as good as externally hosted ones will require a single 16GB GPU and access to basically all of the RAM on the GPU. That's not something people running Firefox on a $500 laptop with 8 or 16GB of total system RAM and a CPU-integrated basic graphics system have to spare.
No it doesn't. TODAY you can run with good performance a local LLM entirely on CPU that can do translation, summaries, search assistance, and all the things mentioned here.
On a 8GB system with a SSD it might be a bit slow due to having to keep experts on disk, but perfectly usable. On fast systems it may well be faster than the network round trip for many tasks.
> No it doesn't. TODAY you can run with good performance a local LLM entirely on CPU that can do translation, summaries, search assistance, and all the things mentioned here.
What specific LLM and how much RAM does it take up to load? How does it run on somebody's 8GB RAM $600 laptop they bought from Best Buy? What pp/s and token/s rate on that hardware?
How much ram has a fair amount of flexibility since MoE can be kept on flash and swapped in at a performance cost, and depending on how small a context you can use.
Well the linked thread mention 14700KF which is a 400$ processor and RTX 2070.
Beside its mentioned 5 or 6 gb of ram.
People run Firefox on pentium with less than 8gb of ram, just 4 or 5 years ago 4gb computer were still extremely common. Even Microsoft sold its surface devices with 4gb of ram, I was dumb enough to buy one.
To give a counterpoint, you could disable it on those low end devices.
But then people would complain about the size, like what happened with Google Chrome embedded llm.
I think there wasn’t any bad choice, mistral was the safest.
My post was an earlier post in this thread. I give figures for pure cpu, not using any GPU. The RTX 2070 figure is a separate figure (with rather insane speeds for that model!). You could contribute to the discussion by grabbing some numbers on lower power hardware if you have some available.
Though low memory devices like 4-8gb probably does need specialized inference to give fair numbers.
> you could disable it on those low end devices
it's not like it does anything if you don't ask for it-- and good thing, because the privacy invasion would be all the worse if it did!
> mistral was the safest
Sending the user's confidential information to third parties while falsely suggesting that it is private can cause them irrecoverable harm.
An alternative is that a feature is slow for some users on slow hardware, and there is a setting to make it faster at the expense of privacy and security that they can switch on. The large body of users that find the default config fast enough have no reason to flip the switch.
> But then people would complain about the size, like what happened with Google Chrome embedded llm.
People or astroturf accounts? :P but having read some of the commentary on it there was some pretty weird takes, like generalized complaints about AI and people thinking google was using their computer to serve other people. Google's business model generally prevents correctly marketing the functionality in any case: it's not like they're going to make a proper pitch for how important it is for your privacy when the rest of their business is centered on hoovering everything up. Mozilla is not so constrained :P
> This is an excellent use case for completely local, small model inference,
How small is this 'small model' inference?
Browsers especially are expected to operate on a very wide number of devices. It makes perfect sense to me to start off with a cloud driven approach and later allow people to configure it to run locally if they have the hardware to do so. Anyone with the hardware to do so is already signaling that they'd have the technical interest to make that change while having the default 'works for everyone' mode be one that doesn't require a powerful local device.
I say this as someone using a 6 year old computer and only upgraded to 8GB of ram when the latest WoW expansion was released.
I suppose some clarity would have been nice but I try to err on the side of charitable interpretation with these things, especially with how much MDS there tends to be in the hn comment sections. This does just seem to be like a high level announcement of the overall partnership, though some clarity over local versus cloud would have been nice.
But they do link to a page describing the "smart window" which goes into a bit more detail:
Yes, that's one of the pages I wrote my comment about. I did try to interpret it charitably, but there's nothing on that page (everyone read it! It's short!) that simply communicates to users, "we're uploading your browsing history to the cloud".
It has a number of words circling around the topic, words relating to privacy and local storage; and all of them avoid communicating the main point. Some seem to mislead about it. How, for example, can you say
> "They’re stored on your device, not on Mozilla’s servers,"
And never explain in plain English "...but also we need to upload them to Mozilla's (& Mistral's) servers" at any point in the advert?
The first time I read this page, I myself couldn't figure out if it was about local inference or not; and I was actively looking for the answer. That's how opaque this is.
> This is an excellent use case for completely local, small model inference
I remember using a small model for autocomplete, I think it was Qwen 2.5 Coder 3B way back, on a laptop with 32 GB of RAM and a 13th Gen Intel i7 CPU. It made the fans spin up and loaded things so much (Ollama back then) that it was largely unusable for doing software development at the same time. I don't want inference anywhere near my personal computing devices (same in regards to phones, really, I don't want the power usage that'll just help degrade the battery faster). Also most small models just really suck, unless you have trained them on a very specific task, but then they suck at everything that is not that task.
Then again sending browser history online also sucks, unless you're into that kind of thing.
Not everyone uses laptops. I can run Qwen 3.8 27B (which is a REALLY capable model) in the background coding for me while I'm simultaneously browsing the web and playing VALORANT without any performance impact, and that's on a 6 year old GPU.
This is ridiculous. Not only does Mozilla support BYOM with Smart Window, but do we seriously expect your average user to be able to locally run it with ease?
Imagine for a moment the complaints that would arise the very next day because their browser is now revving up their laptop fans and hoarding precious RAM.
Local models aren't yet light enough to handle the required processing at the lowest end, so Mozilla chose the most responsible stewards they could find. There is no news here.
zdr is good, really confidential inference with attestation would be better. the same as a confidential vm TEE which confirms the integrity and privacy.
there are good providers available, i think it should become the standard.
You can disable this in about:config by modifying extensions.webextensions.restrictedDomains. It's probably there to help prevent malicious extensions from hijacking official browser support. I reckon many people here don't need it.
You are choosing a really bad example to make your point. Not letting extensions operate on the actual site where you install the extensions isn't being "comically evil" by any means.
This has always been the case for years, for every browser. No idea why you say "now".
Any browser providing extensions, not just Firefox, will at least won't allow extensions to modify their own extensions providing page. Otherwise, any malicious extension can infiltrate and trick users to download modified extensions from the browser's own extension page.
Come on, man, I’m going to choose you’re just Posting saying shit like this.
Because it’s useful. Because almost nobody has the hardware to run a local model with nearly reasonable performance (in intelligence + speed + RAM usage).
Not everything is perfect. At least they’re building things.
The strategies of Google and Apple, regarding how to provide a LLM, seam to disagree with you.
Gemini run on a potato and Apple is local first.
So, you may actually have very good performance with local model. Just not yet on *every* device. So the Mozilla strategy here feel very reasonable. A Cloud provider specialised in local models, to be able to switch once local models will be quick enough on most devices.
Would be very cool if you could type a long query and the model would just build an advanced google search query using what you typed. You could even ship a tiny model in the browser itself that does that.
Something like: "blog posts which talk about using sending emails to yourself as a notes taking mechanism".
The output could look like: "note taking" OR "notes" ("emailing yourself" OR "email to myself" OR "email yourself") (inurl:blog OR inurl:posts OR site:medium.com OR site:substack.com)
I think that's roughly what duck.ai is doing? If I type that in there I see it generating some search queries and then generating a list of links based on the different result pages.
There is. By default it tries to obey your query. But if that query returns no results, only then it makes a best-effort attempt to return something vaguely related.
Yes, and IME it also is true that "online search" means in fact google.com. And there's no distinction for search results and reading the google AI summary.
It's interesting to see Firefox attempt to create slightly more privacy-focused cloud inference infra (assuming you can trust that they adhere to their own policies and don't have bugs, and that their partners adhere to their contractual obligations and don't have bugs). Unfortunately that's still a lot of trust that's pretty much impossible for the end-user to verify, but at least it's something? Arguably it's better than the alternative of directly trusting AI companies that have shown absolutely no regard for other laws like copyright, but the marketing of this as ultra-private still leaves a bad taste. I wish this announcement clarified the cloud aspects directly instead of needing to browse through two layers of links to find how it actually works on a privacy notice.
I don't think it is comparable. Nowhere in the news they say it is offline but IMO Gemini Nano is offline (?)
I'm not a Google fan but I never understood the outrage as it is a local model ... sure, you have to trust them, but if you use their browser you already do.
I mean encryption is a real thing (Edward Snowden sometimes would simply say "Encryption works" in online Q and As) and there are ways to do zero knowledge data crunching Eleven while interacting with servers. Kim DotCom famously launched Mega as a zero knowledge file host after his legal issues with with megaupload, and lots of apps and services across the internet have been doing the version of this even with online services in the ensuing decade and a half. For instance I use StandardNotes which is online but encrypted.
I have no idea the extent to which any of that speaks to the Mozilla thing in this context, and local is preferable if that's achievable. And of course you should always do due diligence on what exactly is and isn't being promised when it comes to sending your data to the cloud, it certainly is possible to do that in privacy preserving ways. It's not a doubt about the technology so much as a question of institutional credibility.
I tried adding my own endpoint, but the save button was still greyed out so it did not work. Maybe the feature is not fully implemented yet? In about:config, I could add the endpoint and api key, but not the model name.
This sounds great, will have to try it out. Interested in the business value they gain from this, as this appears to be free.
I guess for mistral it's great marketing material.
Edit: Quite interested in the differences of local optimizations... that does allow for specific sensorship, but alas that's already possible in a more direct, reliable way.
I think that beyond marketing this is also a way to collect "AI browsing" trajectories. Not an expert but I think this is the kind of great quality data that can be useful when training LLMs.
These are two companies that seem to be slipping in ethics recently. I'd prefer this over a chrome and openai offering, but I no longer trust these two to necessarily do the right thing regarding privacy
"...working together to bring Mistal’s scientific innovations to consumers around the world"
A quick proof-read wouldn't have gone amiss here.
I also fail to see how privacy is enhanced by sending private browsing data to a cloud service which then forwards it to an LLM, in addition to storing 'memories' and dredging through my browser history.
Why would it not be private? Yes data is sent to the servers but the trust that you must have that they don't keep the data is not different from the trust you must have in your password manager or encrypted email or backup storage provider.
All previous similar issues got hand waved with "Mozilla has to make money somehow".
I am afraid one day remaining users will decide that Chromium based browsers provide a better experience in most use cases, and now have same privacy guarantees as Firefox. Looking at market share maybe that had already happened and pragmatism won over ideology, and we will all suffer the consequences of such monoculture.
These comments are so annoying and they pop up on literally any news Mozilla releases. Like… ever.
No, Firefox is not malware because it includes an OPTIONAL feature. Yes, chrome is 100x worse than Firefox when it comes to data privacy. Yes, Firefox works in 99.9999% of websites. Yes, the performance of Firefox is about the same as chrome.
Nobody who seriously gives even a single fuck about privacy is going to be dropping Firefox because of this. Because that’s stupid.
The only conclusion, then, is that people who are dropping Firefox because of this don’t care about privacy at all, and just want an excuse to use chrome. But I guess saying “well I like chrome” is too much, so we just lie instead.
It’s fine to use chrome. It IS NOT fine to lie through your teeth and claim you use chrome because Firefox is just as bad for privacy.
Because it’s not, and not a single person would drop Firefox for that reason. You just want to use chrome. So do that, and stop shitting on Firefox as if Mozilla implementing 0.01% of the malware chrome has is the end of the world.
For me the draw is that Waterfox has partnerships in order to provide funding so if/when Mozilla truly loses the thread (which some may argue they have) WF can keep going on its own, whereas the Librewolf devs have stated that if push comes to shove they can't maintain an entire browser themselves.
Mozilla lost the tech race and with it a lot of users. Firing a lot of their talented developers didn't help. Now Mozilla is giving up the only reason why they kept a core audience of privacy focused users by sending our browsing data to third party servers for inference just at the time when small local models are getting good enough for many/most use cases like summarization or translation. Chrome provides an API for local LLMs. Apple is trying to run as much interference locally as possible. Mozilla is worse than that. Claiming private browsing while sending data to third parties.
I don't care what their data retention policies are, I don't want my data to be sent to others full stop.
Private means it's mine, it's under my control. Handing it to third parties is not private.
AI are excellent at ingesting tons of data and getting something relevant from what you asked.
I ask it about linux behavior and it will dig through hundreds of patches & mails and will return me informations sourced with links to the patches, mails containing it's affirmations.
Didn’t OpenAI tried to build even a new AI browser and failed? I honestly don’t need AI in my browser and would not let an agent to access my login credentials/cookies/saved payment methods for various sites.
It's pretty nice for tedious tasks. I tried to have it look through job listings based on my criteria. It burned through my 5h Claude quota, but what I got was pretty impressive.
I agree on the privacy concerns but this could be useful with a local model, I think.
For me it’s not only privacy, I just don’t trust an agent to use my credentials without doing any destructive task because unlike an MCP, I cannot enforce any permission model on the browser
Not sure if it's a locally run model or in the cloud. If it's the former they don't explain the size on disk and more importantly on memory. I know that google is using local models on the pixel phones, so it's doable. If it's the later, the "Mozilla don't keep the data" is hypocritical and I'm perplexed by the cost associated.
In any case as a french, I prefer my personnal data be exploited by a french company :D
Edit: after reading the comments, it is not a local AI so "Mozilla don't keeep the data and don't train on them" is really hypocritical if they cannot confirm that it's the same on mistral side. And there is still the question of the cost.
I really do not get the obsession with translation, particularly bad translations and translating by default. Most people who browse websites in languages other than the official local one are fluent in that other language.
Google Chrome keeps insisting that Spanish is Galician. PowerPoint insists on changing the spell checker back to the wrong language and ignores setting the language for the whole presentation. On Linux, ChatGPT insists it must enable the spell checker in the local system language, so everything is always red.
Privacy aside, a lot of software seems to be built assuming people cannot be multilingual.
I don't know many people, but I do know someone that relies on machine translation to browse the English web. Remember "most" is not "all." The web should be accessible to every single person.
this is chatgpt, web gui from mistral built-in Firefox, not for tech savvy users. I think anyone that will complain about this knows very well that you can just give your harness control over firefox through puppeteer, which is much more powerful. could firefox itself become a harness and speak openai api responses? maybe yeah, but why?
Remember when it was the fastest thing around and simple?
Now it's an also ran. AI doesn't help here. Focus on what made it great in the beginning, make it the most performant browser around. Put Fable on the case if need be. Do the browser thing well!
Cloud inferrance is not private: These providers, even if they don't get hacked, have bribed employees, or outright lie, cannot and will not resist a subpoena or similar.
Mozilla is acting unethically by obfuscating what's on offer here and overstating the privacy and security properties that can be achieved.
It's doubly unfortunate because simple translations are completely reasonable to do with cpu inference.
Another element of these "claimed to be private but can't keep its promises" is how caustic it is to the trust we have for our friends, family, and business associates.
People here can be privacy aware and well informed and avoid these data slurps-- perhaps go dig up the hidden settings to completely disable it or add firewalls so an errant keypress won't upload your browsing history. That's good.
But anything I share with another person or put on some webpage for another-- perhaps highly trusted person like a doctor or lawyer-- is exposed to them uploading it perhaps completely unwittingly (e.g. they fell for the exaggerated privacy claims) or due to an innocent misclick.
Normalizing privacy fails like this undermines the ability of even the most informed and aware people to opt-out.
So now mistral has partnership with browser + user logs from Mozilla.
Does it means that they are trying to build their own search index?
Because it’s kinda huge inference investment in exchange for what?
Any other ideas why they need this?
I remember the days when people justified Google Chrome's maximalist RAM usage by saying it was a good thing because it was allocating what the program actually needed to run quickly and efficiently, and it was better to grab what it actually needed then to operate inefficiently while using fewer system resources.
For a while that was even the majority view, but it never quite made sense to me.
What is Mistral bringing to the table for Firefox? There is nothing "open" about this in any way.
FF needs to either profile itself as the no-ai-by-default browser, or it needs to just have OpenAI/Anthropic/Mistral/DeepSeek bid for the default spot - like they do with Google search.
Maybe FF believes there is a group of users who are still on the fence about using FF - until they can pitch them a first-class build-in AI story that goes with the anti-establishment vibe?
I somehow doubt that's the pitch & potential userbase they should be focussing on.
This is an excellent use case for completely local, small model inference, yet for inexplicable reasons Mozilla wants to normalize uploading your entire private browsing history to a cloud.
These (Mistral's and Mozilla's) marketing pages aren't candid enough to clearly explain the difference between local and cloud inference, and that they're asking you to consent to enabling the latter. I'd call that the bare minimum of ethics. If you're in a position of authority over your less technical users, you are ethically obligated to give them the full picture of what are you doing and why they should consent.
(Aside to any Mozilla people who might be reading HN, your page here[0] has an oversight, it advertises this model as "Mistral Small 4" but the hyperlink is to OpenAI's model card for gpt-oss-120b).
[0] https://support.mozilla.org/en-US/kb/smart-window-models
I am not trying to defend Mozilla doing this, and I don't support sending data to cloud based services like this in a way that users won't understand.
But I also think that the state of the art in small LLM and user device capabilities aren't there yet to put a "good enough to be actually useful" local-only LLM as a prepackaged thing in a mass market distributed browser.
You don't want a browser that takes 10GB of extra RAM (on top of the memory hog that is having just 3 or 4 complex tabs open on its own already) and pegs your CPU at 99% usage for minutes at a time. And not in an era when mass market consumer laptops are still commonly 8GB or 16GB of total system RAM. Many of those with integrated-into-CPU onboard graphics (eg: not a gaming laptop with a discrete GPU on the PCI-E bus).
It'll be a catastrophe for laptop battery life, among other resource use problems. And an LLM that fits in under 8 to 10GB of RAM for CPU-only inference is not going to be nearly as capable as an off-device inference system.
I wish they had just done this with a very clear up front opt in (not enabled by default) thing that explains what Mistral is, that it's not some big American cloud company but a relatively small startup in France, and that your prompts/LLM interactions will go to their servers. And some documentation on how it will be handled/stored in a supposedly trustworthy manner.
Well, it depends on the task, doesn't it? "running shoes I looked at last week" / "Here's what I found in your browsing history:" doesn't need a 119 billion parameter frontier model; it's a RAG problem for the 0.6 B embedding models. That's an example Mozilla offers. Presumably to explain to their users why it's essential they hand over their last week's browsing history for this convenience (but it isn't! Hardly for that!)
I feel it's wrong to tell users that it's important and normal to relinquish all control of their—extremely personal—life history, in bulk, in plaintext, to strangers.
I agree wholeheartedly that remote server inference is super useful, and that local inference falls far short on many tasks. (I have no objection at all to Mozilla providing a cloud inference feature).
What I don't buy is that we must ask users to redraw their personal boundaries so that their most intimate life details, and remote frontier-model inference, overlap. They do not need to overlap.
You can accomplish a lot with private local inference with the smallest of models; and you can accomplish a lot on remote servers which aren't privy to everything. If some convenience is lost by not combining the two, well, so be it. I'm sure most people would agree, if all of this was laid out plainly.
You’re taking their shoe example way too literally. It’s just showing they’re catering to the average user, not an engineer. And tons of average Joe users now have very high expectations of how intelligent a LLM is because they’ve interacted with ChatGPT and the like. If there’s a super tiny model that is too dumb to do anything beyond shoe searching and they have to switch to Gemini and google AI search for anything more complex, this whole launch would be an immediate failure.
I'm of the opinion that local inference should be done to the greatest extent that is realistically possible, at the earlier time that the hardware/average user platform is capable of doing so. I personally spend a fair bit on kWh extra in my home electrical bill monthly for having a good sized chunk of local inference ability in my house, but that's not a common thing yet.
If mozilla is doing things to send users down the path of doing this externally, they need to be much more upfront and transparent with the users about where their data is going, and not bury it in some terms/conditions that only nerds will hunt for.
> "running shoes I looked at last week" / "Here's what I found in your browsing history:"
Does that need an LLM at all?
Not really, but it's probably easier to make it on top of LLM than to make specially-purposed tool for it, if we talking in terms of time-to-market effort.
SQLite FTS could have done this a decade ago. We've had good local search capabilities for two decades and they either been underused or abandoned (e.g. Google Desktop, Yahoo! Search). This may be a reasonable projection of where AI is headed. You can do a lot locally but there is too much incentive to centralize around cloud infrastructure, then the privacy concerns make that prohibitive and we end up with what feels like a false choice of cloud or bust.
Maybe it's LLM hype that will bring more powerful capabilities to the desktop?
I think classification and clusterisation is more important than search per se. "Where did I see that article about that weird psychological effect where people remember things they haven't seen" is not resolved by direct search, but can be - with some work - helped by language models and NLP. One doesn't need full blown frontier LLM for that, but bare word indexing would probably not do either.
Firefox is already using small local models for a few things, like tab grouping and link previews. I assume that this is for cases where a reasonably sized model isn't adequate.
> Firefox is already using small local models for a few things, like tab grouping and link previews
Which I immediately disabled alongside their "improved" Nova design, cause it doesn't improve productivity. These are all distractions from web browsing. If they want to improve anything, they should work on their abysmal spell checker, but that's not shiny enough.
Let's say the goal is doing something for work, so accuracy is important, and you prefer for it to be a nice reading experience, like a good translation.
As long as cloud model are somewhat better in these things, it's good that users would have the option to use cloud models.
Dont use this for work unless it has been verified by your corporate security and legal teams. Sharing business data with a company with which you dont have a enterprise agreement can land you in serious trouble.
I'll ding Mozilla directly then - Since your "we can't do this yet" approach is actually the literal thing Chrome is shipping...
https://developer.chrome.com/docs/ai/built-in/overview?_gl=1...
But the built in AI model in Chrome is 4GB of RAM, 4GB of disk space, and is still only used in a couple places
> that it's not some big American cloud company but a relatively small startup in France
How it's any better? Small companies can be bought by big companies. French government is as capable of trampling over their citizen's privacy the moment they feel they need it as US government is. And putting a squeeze on a small startup is way easier than on a major cloud company (not that either is particularly hard). Also, OpenAI used to be an idealistic non-profit one day too, then it started to smell trillions and all that went of of the window.
> is not going to be nearly as capable as an off-device inference system.
I rarely need PhD-level research into my browsing history. I'm not going to solve millennium problems on my bookmarks. The tasks that I will realistically need are well within capacity of most very basic local models. Maybe they'd be a bit slower, who cares.
Let's not pretend that EU data laws are not far more stringent and actually enforced than whatever the US has. Like sure, the government with a jury decision may acquire some person's data. But it's not getting sold to whichever-random-company-offered-the-highest.
Also, some shady TOS is also not enough to override the law in the EU, you have much more protection as a user here.
> But it's not getting sold to whichever-random-company-offered-the-highest.
Nothing in GDPR prevents transfer or selling of personal data. True, there are hoops to be jumped to do that, both in terms of consent and documentation, but let's not pretend a lot of people would read the welcome banner and refuse to interact with a service because its legalese says "we'll sell you data and if you don't want us to, go away". Yes, selling the data becomes more expensive because you need to hire the lawyers to produce those welcome pages and regulatory compliance paperwork, but large corps have enough money for lawyers.
> Also, some shady TOS is also not enough to override the law in the EU
It doesn't need to override anything, as the law does not prohibit data collection or transfer. It only describes the hoops that need to be jumped to achieve it.
It certainly does. You need to weigh the users privacy with your interests and can’t just put them aside. Noncompliance leads to quite large fines. You can easily find examples of fines given.
The DPA’s are short on capacity, sure, and large corps will try to fight the fines in court. But to describe it as “just hoops whilst allowing everything” is unfair.
> You need to weigh the users privacy with your interests and can’t just put them aside.
Who said anything about putting them aside? There would be a process and highly paid lawyers confirming that selling user data is only for the user's ultimate benefit, as it allows to provide awesome services to the users, and it all will be outlined in a 100-page privacy policy which you will be sure to read on each site you use, wouldn't you?
> But to describe it as “just hoops whilst allowing everything” is unfair.
Can you quote me the place where GDPR prohibits this? Not says something like "weigh user privacy" and "take adequate measures" and so on - which can always be resolved as "we weighed carefully and we took measures and we decided selling the data was awesome and users love it" - but explicitly and unambiguously prohibits the practice? If you don't find it - that description is exactly what it is.
>How it's any better? Small companies can be bought by big companies. French government is as capable of trampling over their citizen's privacy the moment they feel they need it as US government is.
It's better for Europeans because the company in question is European and so will not export their data to New Jersey. And because any success it has will presumably better benefit France and Europe.
> It's better for Europeans because the company in question is European and so will not export their data to New Jersey
There aren't many government institutions in New Jersey that would take a lot of interest in anything Frenchmen are doing. There are a lot of institutions in France that would be interested in anything Frenchmen are doing, if it contradicts what the French government wants to be happening. The main threat to citizen's privacy always comes from the government closest to them, the government overseas has its own citizens to worry about and pays much less attention to foreign citizens on foreign land. There could be exceptions, true, but as a rule, if you look into New Jersey, you'd sooner find mass surveillance of American citizens than mass surveillance of the French.
Same goes for commercial interests. If I want to run targeted ads in New Jersey, I want to have profiles of New Jersey people, not French people. So buying data from France in New Jersey would not be a routine occurrence, but buying local data would be.
> How it's any better? Small companies can be bought by big companies.
This. I've come to view a startup as a company without a business model, doing everything to get acquired by a mega corp that will finally squeeze the juice out of the userbase
(The lack of a viable business model applies to some mega corps too)
>But I also think that the state of the art in small LLM and user device capabilities aren't there yet to put a "good enough to be actually useful" local-only LLM
so then don't add it in as highly advertised feature until it is. doing things right and living up to your core values is a lot to expect from businesses these days but, at minimum, a non-profit foundation should be able to live up to these goals, yes?
But apple said their AI only falls back to the private cloud when it has to? Just kidding, near every request needs to fallback, because a phone can't actually run a real LLM, no matter how many "neural cores" it has..
Agree, advocating for Firefox developing features only for rich hobbyists(people who can afford large RAM and GPU) is absurd.
>advocating for Firefox developing features only for rich hobbyists
?:
>marketing pages aren't candid enough to clearly explain
The only people having an expectation of translations being done locally is exactly the nerds that keep whining that it's not using a local model. Every single normal person, when presented with a "translate" button either know it's going online, or don't care about it.
Begging the purists to run away from Firefox at this point so they can stop wasting everyone's time. Your demands for examplarity and whining about money not going ONLY to firefox and jerking yourselves on Servo was not enough, now you want to restrict the browser to owners of an RTX5080 if they want to use it?
> only people having an expectation of translations being done locally is exactly the nerds that keep whining that it's not using a local model
This is an untested assumption in Silicon Valley. I suspect Apple is going to eat a lot of folks’ lunches.
You realize that people that don't care won't even know what Firefox is? Being a niche player and scaring away niche people is the most stupid strategy possible.
Being a small player and throwing away any hope of expanding your user base by building for people who will complain no matter what you do is, well, something!
neglecting existing customers in an attempt to gain larger market share is certainly a play
Ah yes, the customers of Mozilla, who are certainly paying users and not just people whining and saying BACK IN MY DAYS IN THE LAST MILLENIUM IT WAS DIFFERENT.
Get over it gramps, things are changing. Mozilla spent ten years trying to mostly appeal to these people by staying very conservative, and it got them nothing but abuse and harassment from people not realising that staying stuck in the past is a death sentence for Mozilla.
You want to save Mozilla, burn down Google first. The ecosystem in which they are evolving isn't driven by them.
I solemnly proclaim I would pay reasonable money for a browser (Mozilla or otherwise) if their maker would truly try to make the best browser possible and succeed at least partially at it, instead of serving a thousand of agendas none of which is making the best browser possible. The sole reason why I am not Mozilla's customer is exactly that - because they do not see me as their potential customer and don't care about my needs. Of course, they are free to do that but the result is people like me aren't giving them their money. So they have to go to Google.
like I said, neglecting existing customers in an attempt to gain larger market share is certainly a play
the rest of your post seems to hinge on some judgement you think I made about that
> You want to save Mozilla, burn down Google first.
Uhhh... 85% of mozilla's revenue comes from Google via the search deal. In a theoretical scenario where a genie waved a magic wand tomorrow and google went "poof", disappearing, mozilla would be in a dire emergency for lack of revenue.
It would. But Mozilla has zero long term survival chance in the current ecosystem. The economics are just not there, there cannot be more than one big browser in the way this system works.
[flagged]
Whoa - you can't attack other users like this on HN, no matter how wrong someone is or you feel they are.
We ban accounts that post like this, as you surely know. We don't want to ban you, so please don't post like this!
https://news.ycombinator.com/newsguidelines.html
I do not fundamentally disagree with you but it's also an extremely well known phenomenon that average non tech users, in the aggregate of millions of people, will click almost any "yes/I agree/continue/Next" step on a software installer or new user sign up workflow for anything, without reading the ToS. People blithly sign up for all sorts of cloud based things and services without understanding their full ramifications all the time.
I also wonder at the specific level of aggression and the tone of your comment which does not seem to be an appropriate response to that person's specific comment.
People such as you are describing and rightfully criticizing are knowingly taking advantage of that. Indeed it's how a lot of malware gets installed too.
The person you're responding to is pointing out that a lot of people at the surface level do only appear to care about the results. They put something into google translate, it works, they gets results they are pleased with, they don't put a lot of thought into the fact that the data is going to an external service. That's not an inaccurate description of how a lot of people use their computers these days. Look at how many people signed up for ChatGPT accounts and put the chatgpt app on their phones and talk to it all the time. That's the level of critical thinking a lot of non tech users have about their personal data.
The fact that people will click yes/agree/OK on almost anything is how Windows computers got Bonzi Buddy installed on them back in the day, and now it's continued into the cloud-everything era.
https://geekhack.org/index.php?action=dlattach;topic=21140.0...
> it's also an extremely well known phenomenon that average non tech users, in the aggregate of millions of people, will click almost any "yes/I agree/continue/Next" step on a software installer or new user sign up workflow for anything, without reading the ToS. People blithly sign up for all sorts of cloud based things and services without understanding their full ramifications all the time.
I would put a lot of that down to learned helplessness.
A lot of non-tech people tell me that "they already know" and that it is impossible to avoid. I have been told I am naive to think its possible to keep data private.
It's almost as if outside the tech bubbles most people don't understand the tech or implications. HN posters who live and breath in this space are so smug to pretend this esoteric knowledge about networking identifiers or fingerprinting is simple and that anyone can get it. These are the same people who will throw their hands up on any other esoteric topic and give up on it equally. For example, tech people so easily decry how "stupid" the law and legal jurisprudence is (because obviously if they don't understand it, it must be bullshit) but see zero irony in being smug pretentious assholes to regular people who didn't happen to major in computer science.
We shouldn't all have to be PhDs in law and jurisprudence to trust a legal system to look after our interests just like we shouldn't all have to be computer scientists to trust that technical stewards are not exploiting people's lack of knowledge.
Sick of this mentality. Oh these people are sheep idiots who don't understand the EULA euphemisms of "improved user experience " means adtech and behavioural tracking.
I think youre right. The least techy people I know where the first to ditch Mozilla when they started to see ads in Firefox.
Can you please not go on enraged tirades on HN? You've posted several fulminatory comments in this thread alone:
https://news.ycombinator.com/item?id=49729021
https://news.ycombinator.com/item?id=49728539
https://news.ycombinator.com/item?id=49728246
https://news.ycombinator.com/item?id=49728175
and at least one more in another thread today. This is not what HN is for and it destroys what it is for. You may not owe Mozilla better, but you owe HN better if you want to participate here. This is only a place where people want to participate becuase enough people make the effort to raise the standards rather than drag them down. https://news.ycombinator.com/newsguidelines.html
I love this site
PREACH
That's a lot of words and assumptions, when a single check of my posting history would show that I despise the HN bros as well. Jumping on a tangent about libertarianism when you could have simply called them retarded and saved a lot of time.
Anyways, no, I'm talking about the average person, the public worker, the person that thinks the internet is the funnily named Safari app, the elderly: they give zero fucks about it going to some service online. They used to search for Google translate before, whether or not it goes on someone's server, they do. not. care. You're not going to win them over with "it runs on your device". Their device is a crappy laptop that barely runs excel, and if they can offload computing, they will.
They're advocating for a CHOICE and for the difference to be explained.
Have you tried Ling-3.0-tiny? It runs fine on CPU-- on a 14700KF gets 40tg/s and 250pp/s and on a ordinary gpu (RTX 4070) does over 200tg/s with no MTP and 7185pp/s. (my figures are Q8, though presumably a good Q4 would be faster)
It's certainly not as capable as something that needs a high memory gpu for quick performance, but I was quite impressed with it for what it is.
(and fwiw, I had it translate your last paragraph to German, then used google translate back to english: "I wish they had handled this clearly and transparently via an opt-in mechanism—not enabled by default—that explains what Mistral is (not a major American cloud company, but a relatively small French startup) and that your prompts and LLM activities are sent to their servers. I also wish there were documentation explaining how the data is handled and stored in a way that inspires trust.").
How much RAM does it take up in total? I'll have to give that a try on one of my test systems. Looking at a somewhat randomly chose GGUF quantization of it, looks like just under 5GB on disk in Q4, so RAM usage somewhere around 5-6GB?
https://huggingface.co/bartowski/Ling-3.0-tiny-GGUF
That sounds about right for Q4.
it's an extremely sparse MOE, so there is some odds of acceptable performance using a smaller in-memory cache and the rest on flash. ... I don't have a setup to test that right now.
(Of course, if translation is all you want much smaller models will work. Ling-tiny can do summarization, dom manipulation, scripting, etc. too).
Andreessen Horowitz led Mistral's €385M Series A in December 2023.
Is that like 2-4 RTX9000 cards price rn?
> small LLM
Read this again, slowly.
> You don't want a browser that takes 10GB of extra RAM (on top of the memory hog that is having just 3 or 4 complex tabs open on its own already) and pegs your CPU at 99% usage for minutes at a time.
But I do, especially when the choice is either having it locally, remotely, or not at all. I also indeed do nit my CPU used for such, but GPU. I could even run a 8 GB model on a remote (but still local network, on-prem) NPU.
There's one caveat though: if you are gaming and browsing.
> But I also think that the state of the art in small LLM and user device capabilities aren't there yet to put a "good enough to be actually useful" local-only LLM as a prepackaged thing in a mass market distributed browser.
Then just allow it to be enabled on high end devices? But it must be local only. As hardware advances and people upgrade, more people will be able to turn on the feature.
If the thing you wan to do is not possible without totally compromising your ideals, maybe don't do it ?
> I am not trying to defend Mozilla doing this
Well, you kind of are though.
> I wish they had just done this with a very clear up front opt in
If a local model is not realistic, then this should not have even been an in-your-face opt-in, but at most some add-on.
Of course, their telemetry isn't even opt-out, so even the opt-out for the Mistral thing is kind of disingenuous on their part, since they get a bunch of information from us in other ways.
(sigh) Ah, Mozilla has gone down such a dark path over the years. Too bad.
I'm worried that the middle could fall out of the computing market across the board. If you can afford to keep up with the upgrade treadmill, you'll get private, local inference capabilities. If you can't afford to stay on the treadmill, you'll be stuck with whatever cloudshit malware Silicon Valley wants to foist on you.
I acknowledge that this is already the case, to some extent. The cheapest laptops at Best Buy are crammed with the most preinstalled malware. That's been the case for, what, 25 years? But you've always been able to wipe that cheap laptop and make it into a much more capable, trustworthy machine.
Well, assuming LLMs do become a pervasive part of the computing experience, what happens to the cheap laptops? Do all computers get more expensive to accommodate local inference? Does the rift between the everyday user's experience and the savvy user's experience grow even wider than it already is? Neither outcome seems good for the average joe who just needs to check his email.
Remember how in like 1999/2000 Sun was trying to predict that everyone's computer would be some form of thin terminal in the future? Turns out they were very wrong on the part about it running on Sun server back-end infrastructure, but that same general purpose has now been accomplished through other methods where a lot of people do basically EVERYTHING inside a web browser tab to some external cloud service.
Now add the need for external inference because very few random consumers are going to buy a $3000 laptop when they can get the $600 laptop at Best Buy, and that trend further escalates.
This has always been the case, forever. You have to pay for a product or service. How you do so can be with cash or your data/body/vote/eyeballs/indirect discretionary purchases.
The amount of work that can be done funded by foundations and free work is nowhere close to what people want.
Oh give me a break. This argument that people's objections to advertising comes from some Pollyanna naivety over things being free is nonsense. Tell me the last time you have ever seen a company be upfront and explicitly offer a free tier where they tell you upfront exactly what they are harvesting about you and for what purpose (and no, "improving user experience" isn't being upfront) but also offer you a paid version where they explicitly promise not to do that.
"If the product is free then you are the product" is supposed to be a cautionary observation, not an axiomatic proscription.
Does anyone actually trust companies that offer paid services to not harvest their data? Whenever I see this trope I think to myself that whatever lack of regulation, oversight and enforcement lead to that being okay would equally allow for them to both take my money and harvest all my of data anyways.
This is akin to people who criticise socialised medicine or services and pejoratively characterize supporters as just wanting "free stuff", as if the concept of collective payment is naive or something
Every ounce of RAM and spare cycle should be used.
[dead]
[flagged]
The privacy policy for Smart Window[1] indicates that prompts are not only sent to an LLM provider (Mistral in this case), but also to Mozilla servers first.
Where they do use an on-device model is for the initial query classification: [1] https://www.mozilla.org/en-US/privacy/smart-window/This sentence is eye-opening:
Privacy-minded users who want to try this feature would need to create a totally separate browser profile to avoid data leaks.You can control this in the settings:
The real question here is what is the default? Is it opt-in or opt-out?
[flagged]
Please don't fulminate on HN. The guidelines make it clear we're trying for something better here. https://news.ycombinator.com/newsguidelines.html
Some years ago Firefox was the go to browser if you wanted to have "some" privacy in your browsing (together with uBlock origin). With news like this I m really wondering if my views are outdated and I need to to some good researching on maintaining some privacy in what I m browsing.
Mozilla has been in a bit of a recursive feedback loop death spiral for years now of: Browser loses market share -> try some weird thing -> very few people like or use it -> browser loses market share -> management says "oh shit we're losing market share we better try some weird thing"
I say this as a person that uses firefox with ublock origin 99.5% of the time. It's better in my opinion than Chrome. I can at least easily turn off the enabled by default crap features. Like the advertising and sponsored news links on the default new tab page.
This is the confidently incorrect narrative that gets repeated in the comment sections practically every time Mozilla is mentioned but it's every bit as incorrect now as it's been all the previous times.
So here we go again for the millionth time: the big losses of Mozilla market share were approximately during the 2010s. The era of side bets on unique features is approximately the 2020s. The unique features didn't retroactively cause the market share losses of the 2010s.
Moreover, telling the market share story in terms of specific browser features misses the elephant in the room, which is that Google, with the world's most visited page, and a browser that's the installed default on over a billion devices, grew it's market share with a combination of web visability and dominance over the most used mobile platform. Mozilla could triple their budget and have the world's best browser experience, but it wouldn't make much of a dent against distribution defaults.
I would wager that the impact on market share is driven about 97% by Googles distribution advantage and 3% by aligning with user preferences on features and performance. If being a perfect browser led to market dominance, Opera would have already conquered the world back in 2012, but the economics of building a browser aren't always friendly to the good guys.
FF's market share crashed because it was painfully slow compared to chrome when it first came out. The marketing helped, but it wouldn't have stuck if it hadn't been better. You can only blame users being too stupid to know better for just so long.
Exactly. Chrome's growth has been driven primarily by monopolistic behavior, and the fact that FF was in a bad spot when Chrome came out.
Firefox is frankly quite good right now, yet they still are losing or not gaining users.
I never said that Chrome and Edge don't have a huge distribution advantage due to being installed by default on peoples' devices. That's also certainly a huge factor in the mass adoption of Chrome as what people consider "the web browser. Or Safari as default browser in MacOS of course.
Why do they keep trying weird things? That makes no sense. People just want a browser.
Because just doing a browser is not working. Market share keeps declining. And Firefox is pretty good. They spend the majority of their resources on Firefox, and always have.
very few people like or use it -> browser loses market share
I very much doubt trying weird things meaningfully changed their market share (aside from UI redesigns)
I very much think it did, because FF once was recommended by us nerds, we installed it for people and praised it. And that made an impact outside of our circles.
Now I still occasionaly install it along with ublock origin because there is no alternative, but I don't praise it anymore (or bother to install it for someone in the first place) - but rather bitch about how they also sneak in advertisement and spyware.
I think it didn't help. Constantly adding or changing features makes it less consistent and predictable. For example several family members were confused by what is "pocket".
They have twice (that I'm aware of) tried rebranding as an advertising company. They are funded by Google (though not _necessarily_ influenced by them). There was an article in LWN by them about (in part) how they can't get enough information about users of Thunderbird so they were suggesting telemetry should be enabled by default and opt out instead of opt in because most people won't opt out.
I think they get a lot of credit because they aren't Google and not enough push back
Here's a crazy idea if you don't want Smart Window to send your data anywhere:
have you tried not using smart window (like you're currently doing)?
The problem is, many of the privacy invading "features" are opt-out, instead of opt-in...
https://privacytests.org/
Those people complain about everything. It's exhausting.
Fyi, you may want to check out Librewolf: https://librewolf.net/
They are and Mozilla is basically malicious org
[dead]
I can't fathom how they'd consider "Private" something that uploads all the sensitive information in readable form to a third party service.
In fact, If I were asked to explain the opposite of private, I might give this kind of design as an example.
It's more un-fathomable if you recall Mozilla were the people who *pioneered* local browser ML as a privacy feature. Back in 2022—when Chrome was doing language translation by uploading to Google servers, and Mozilla introduced the (actually innovative!) idea of baking small ML models into the browser,
https://news.ycombinator.com/item?id=33792447 ("Firefox Translations: Translate websites in your browser without using the cloud (addons.mozilla.org)")
And look: they were 100% cognizant, then, of how vital local processing is to privacy:
> "...called Project Bergamot. The ultimate goal of this consortium was to build a set of neural machine translation tools that would enable Mozilla to develop a website translation add-on that operates locally, i.e. the engines, language models and in-page translation algorithms would need to reside and be executed entirely in the user’s computer, so none of the data would be sent to the cloud, making it entirely private."
https://blog.mozilla.org/en/mozilla/local-translation-add-on... ("Mozilla releases local machine translation tools as part of Project Bergamot")
Even more un-fathomable, Mozilla's CTO has been writing a substack with the underlying theme of "if the AI isn't running on your hardware, you don't own it" https://newsletter.ownersnotrenters.com/
They're still 100% cognizant! And yet.
You cannot ship on device inference with a real LLM still. Average people freak the hell out if it's even 10% slower than whatever google ships. At least thats the claim for how Firefox lost all their marketshare in the first place (Which was not the case. When everyone was claiming firefox was "slow", it simply was not, as long as you used an ad blocker. Normal people like my dad didn't switch to Chrome. It was installed through a sketchy mechanism and they never noticed)
Meanwhile, if you CAN run these models locally, using your own hardware for inference is supported. It's a setting option.
Google still blasts you to install Chrome at every opportunity if you use their services, and is heavily optimized for Chrome (e.g. Docs, YouTube). That's not really Firefox's fault.
> This is an excellent use case for completely local, small model inference, yet for inexplicable reasons Mozilla wants to normalize uploading your entire private browsing history to a cloud.
I certainly agree that this option would be nice to have, but running firefox without an ad blocker on the modern web will ALREADY spin up my fans just rendering all the garbage, and I have a decent laptop.
You can't expect the median Firefox-running device to be able to run a capable enough local model effectively.
Just gonna go ahead and adopt the phrase "spin up my fans" to describe pissing me off.
> This is an excellent use case for completely local, small model inference
Is it really ? LLM take lot of ram and drain battery. People run Firefox on low end computer.
Indeed. Realistically a 'capable' small local LLM, even one that's definitely not as good as externally hosted ones will require a single 16GB GPU and access to basically all of the RAM on the GPU. That's not something people running Firefox on a $500 laptop with 8 or 16GB of total system RAM and a CPU-integrated basic graphics system have to spare.
No it doesn't. TODAY you can run with good performance a local LLM entirely on CPU that can do translation, summaries, search assistance, and all the things mentioned here.
On a 8GB system with a SSD it might be a bit slow due to having to keep experts on disk, but perfectly usable. On fast systems it may well be faster than the network round trip for many tasks.
> No it doesn't. TODAY you can run with good performance a local LLM entirely on CPU that can do translation, summaries, search assistance, and all the things mentioned here.
What specific LLM and how much RAM does it take up to load? How does it run on somebody's 8GB RAM $600 laptop they bought from Best Buy? What pp/s and token/s rate on that hardware?
https://news.ycombinator.com/item?id=49728887
How much ram has a fair amount of flexibility since MoE can be kept on flash and swapped in at a performance cost, and depending on how small a context you can use.
Well the linked thread mention 14700KF which is a 400$ processor and RTX 2070.
Beside its mentioned 5 or 6 gb of ram.
People run Firefox on pentium with less than 8gb of ram, just 4 or 5 years ago 4gb computer were still extremely common. Even Microsoft sold its surface devices with 4gb of ram, I was dumb enough to buy one.
To give a counterpoint, you could disable it on those low end devices.
But then people would complain about the size, like what happened with Google Chrome embedded llm.
I think there wasn’t any bad choice, mistral was the safest.
My post was an earlier post in this thread. I give figures for pure cpu, not using any GPU. The RTX 2070 figure is a separate figure (with rather insane speeds for that model!). You could contribute to the discussion by grabbing some numbers on lower power hardware if you have some available.
Though low memory devices like 4-8gb probably does need specialized inference to give fair numbers.
> you could disable it on those low end devices
it's not like it does anything if you don't ask for it-- and good thing, because the privacy invasion would be all the worse if it did!
> mistral was the safest
Sending the user's confidential information to third parties while falsely suggesting that it is private can cause them irrecoverable harm.
An alternative is that a feature is slow for some users on slow hardware, and there is a setting to make it faster at the expense of privacy and security that they can switch on. The large body of users that find the default config fast enough have no reason to flip the switch.
> But then people would complain about the size, like what happened with Google Chrome embedded llm.
People or astroturf accounts? :P but having read some of the commentary on it there was some pretty weird takes, like generalized complaints about AI and people thinking google was using their computer to serve other people. Google's business model generally prevents correctly marketing the functionality in any case: it's not like they're going to make a proper pitch for how important it is for your privacy when the rest of their business is centered on hoovering everything up. Mozilla is not so constrained :P
> This is an excellent use case for completely local, small model inference,
How small is this 'small model' inference?
Browsers especially are expected to operate on a very wide number of devices. It makes perfect sense to me to start off with a cloud driven approach and later allow people to configure it to run locally if they have the hardware to do so. Anyone with the hardware to do so is already signaling that they'd have the technical interest to make that change while having the default 'works for everyone' mode be one that doesn't require a powerful local device.
I say this as someone using a 6 year old computer and only upgraded to 8GB of ram when the latest WoW expansion was released.
I suppose some clarity would have been nice but I try to err on the side of charitable interpretation with these things, especially with how much MDS there tends to be in the hn comment sections. This does just seem to be like a high level announcement of the overall partnership, though some clarity over local versus cloud would have been nice.
But they do link to a page describing the "smart window" which goes into a bit more detail:
https://www.firefox.com/en-US/smart-window/?view=update
Yes, that's one of the pages I wrote my comment about. I did try to interpret it charitably, but there's nothing on that page (everyone read it! It's short!) that simply communicates to users, "we're uploading your browsing history to the cloud".
It has a number of words circling around the topic, words relating to privacy and local storage; and all of them avoid communicating the main point. Some seem to mislead about it. How, for example, can you say
> "They’re stored on your device, not on Mozilla’s servers,"
And never explain in plain English "...but also we need to upload them to Mozilla's (& Mistral's) servers" at any point in the advert?
The first time I read this page, I myself couldn't figure out if it was about local inference or not; and I was actively looking for the answer. That's how opaque this is.
> This is an excellent use case for completely local, small model inference
I remember using a small model for autocomplete, I think it was Qwen 2.5 Coder 3B way back, on a laptop with 32 GB of RAM and a 13th Gen Intel i7 CPU. It made the fans spin up and loaded things so much (Ollama back then) that it was largely unusable for doing software development at the same time. I don't want inference anywhere near my personal computing devices (same in regards to phones, really, I don't want the power usage that'll just help degrade the battery faster). Also most small models just really suck, unless you have trained them on a very specific task, but then they suck at everything that is not that task.
Then again sending browser history online also sucks, unless you're into that kind of thing.
Not everyone uses laptops. I can run Qwen 3.8 27B (which is a REALLY capable model) in the background coding for me while I'm simultaneously browsing the web and playing VALORANT without any performance impact, and that's on a 6 year old GPU.
Thanks for mentioning the mistake on the link, it's been reported internally.
The "Gemini 3.1 Flash Lite" also points incorrectly (?) to https://storage.googleapis.com/deepmind-media/Model-Cards/Ge...
> for inexplicable reasons Mozilla wants to normalize uploading your entire private browsing history to a cloud.
Mozilla does a great job of doing all sorts of things that help Google, these days…
the average user has nowhere near enough the capability to run a model fast enough to do useful things
> "...the bare minimum of ethics..."
That is a great phrase.
This is ridiculous. Not only does Mozilla support BYOM with Smart Window, but do we seriously expect your average user to be able to locally run it with ease?
Imagine for a moment the complaints that would arise the very next day because their browser is now revving up their laptop fans and hoarding precious RAM.
Local models aren't yet light enough to handle the required processing at the lowest end, so Mozilla chose the most responsible stewards they could find. There is no news here.
zdr is good, really confidential inference with attestation would be better. the same as a confidential vm TEE which confirms the integrity and privacy.
there are good providers available, i think it should become the standard.
two high quality companies: near.ai tinfoil.sh
Anyone else's extensions not work on support.mozilla.org?
You can disable this in about:config by modifying extensions.webextensions.restrictedDomains. It's probably there to help prevent malicious extensions from hijacking official browser support. I reckon many people here don't need it.
[flagged]
You are choosing a really bad example to make your point. Not letting extensions operate on the actual site where you install the extensions isn't being "comically evil" by any means.
[flagged]
This has always been the case for years, for every browser. No idea why you say "now".
Any browser providing extensions, not just Firefox, will at least won't allow extensions to modify their own extensions providing page. Otherwise, any malicious extension can infiltrate and trick users to download modified extensions from the browser's own extension page.
Comically ignorant.
support.mozilla.org and addons.mozilla.org are distinct sites.
its up to us to make extensions that are so famous that they can be integrated into the browser functionality i guess
Except ... this is not local. It's a remote service. Here is something that is:
Configure them. Open-webui needs a model (a local qwen3.8 does amazing), and you need to link them up. Playwright will just do stuff.https://support.mozilla.org/en-US/kb/smart-window-byom
Also supports Lemonade and Ollama, which are.
But Mozilla is bad and stupid! CEO pockets in a lot of money!!1!
Oohohohohoohohohoho Merry Christmas!!!
> yet for inexplicable reasons
Come on, man, I’m going to choose you’re just Posting saying shit like this.
Because it’s useful. Because almost nobody has the hardware to run a local model with nearly reasonable performance (in intelligence + speed + RAM usage).
Not everything is perfect. At least they’re building things.
The strategies of Google and Apple, regarding how to provide a LLM, seam to disagree with you. Gemini run on a potato and Apple is local first.
So, you may actually have very good performance with local model. Just not yet on *every* device. So the Mozilla strategy here feel very reasonable. A Cloud provider specialised in local models, to be able to switch once local models will be quick enough on most devices.
Would be very cool if you could type a long query and the model would just build an advanced google search query using what you typed. You could even ship a tiny model in the browser itself that does that.
Something like: "blog posts which talk about using sending emails to yourself as a notes taking mechanism".
The output could look like: "note taking" OR "notes" ("emailing yourself" OR "email to myself" OR "email yourself") (inurl:blog OR inurl:posts OR site:medium.com OR site:substack.com)
I think that's roughly what duck.ai is doing? If I type that in there I see it generating some search queries and then generating a list of links based on the different result pages.
That’s a great idea, and would be a much better feature than the unimaginative things Mozilla has been implementing. It’s so disappointing.
There's no such thing as an advanced google search query anymore. The second will give you worse results than the first (without quotes.)
There is. By default it tries to obey your query. But if that query returns no results, only then it makes a best-effort attempt to return something vaguely related.
Well, it would be helpful for scholarly articles.
I feel like much of what Perplexity does can be accomplished that way.
It wouldn't be cool because it would still hand your data to Google.
I assume Google was just a placeholder for a search engine in the example. And you'll surely need some service to search the web.
I don't know about that. When I hear people reference "Google" it seems to me they do mean Google. Similar with "Claude".
We’ve got the whole post to parse, though.
The idea is obviously applicable to other search engines.
People still use “Google” as a synonym for “online search.”
Yes, and IME it also is true that "online search" means in fact google.com. And there's no distinction for search results and reading the google AI summary.
It could anonymize the data, at least.
It's interesting to see Firefox attempt to create slightly more privacy-focused cloud inference infra (assuming you can trust that they adhere to their own policies and don't have bugs, and that their partners adhere to their contractual obligations and don't have bugs). Unfortunately that's still a lot of trust that's pretty much impossible for the end-user to verify, but at least it's something? Arguably it's better than the alternative of directly trusting AI companies that have shown absolutely no regard for other laws like copyright, but the marketing of this as ultra-private still leaves a bad taste. I wish this announcement clarified the cloud aspects directly instead of needing to browse through two layers of links to find how it actually works on a privacy notice.
Seems like what Chrome has with the default built-in Gemini Nano model.
Hey at least "some" news/things from Mistral. Seems like ages ago when they launched vibe-code.
> Powers context-aware search, page summaries, and memory retrieval across browser tabs.
> Live in France and North America, with launches in the UK and Germany planned for later this year.
> Built on a zero data retention policy; conversations are not saved on Mozilla’s servers by default.
> Models are fine-tuned on regional languages, dialects, and local nuances to provide native-feeling responses.
> Extends Mistral’s open-weight AI models directly to consumer web browsing alongside its enterprise offerings.
I don't think it is comparable. Nowhere in the news they say it is offline but IMO Gemini Nano is offline (?)
I'm not a Google fan but I never understood the outrage as it is a local model ... sure, you have to trust them, but if you use their browser you already do.
The outrage about Gemini Nano in Chrome was the surprise extra 4GB download plus their attempt to expose it as a non-standard browser API, IIRC.
> Nowhere in the news they say it is offline
Reading HN, I 100% understood “private” to be marketing copy for “local inference”.
How else could the information be kept private?
I begrudgingly suppose that there is precedent for the usage, like Apple’s Private Relay, but I really don’t think they are comparable.
> conversations aren’t saved on Mozilla’s servers by default, and partners like Mistral agree to zero data retention.
I don't read this as offline ...
People only read the headline, and the headline lies.
> Reading HN, I 100% understood “private” to be marketing copy for “local inference”.
Pretty ironic then that here, Mozilla is specifically labeling their "we send it all to the cloud" service as "private."
So does duckduckgo with their duck.ai. No outrage there.
I mean encryption is a real thing (Edward Snowden sometimes would simply say "Encryption works" in online Q and As) and there are ways to do zero knowledge data crunching Eleven while interacting with servers. Kim DotCom famously launched Mega as a zero knowledge file host after his legal issues with with megaupload, and lots of apps and services across the internet have been doing the version of this even with online services in the ensuing decade and a half. For instance I use StandardNotes which is online but encrypted.
I have no idea the extent to which any of that speaks to the Mozilla thing in this context, and local is preferable if that's achievable. And of course you should always do due diligence on what exactly is and isn't being promised when it comes to sending your data to the cloud, it certainly is possible to do that in privacy preserving ways. It's not a doubt about the technology so much as a question of institutional credibility.
New apple watch markets their new listening feature as private, but they are not happening on device.
>like Apple’s Private Relay
Reading this as “Private Cloud Compute“ (PCC)
(Not iCloud Private Relay, the VPN)
Private here means "trust me, bro".
It’s not obvious but you can use this with your own local (or any) models.
https://support.mozilla.org/en-US/kb/smart-window-byom
I tried adding my own endpoint, but the save button was still greyed out so it did not work. Maybe the feature is not fully implemented yet? In about:config, I could add the endpoint and api key, but not the model name.
Found the reason, HTTP-endpoints has to be localhost or 127.0.0.1, if you want to use other endpoints it needs to be HTTPS.
Keys that needs to be set in about:config
- browser.smartwindow.apiKey
- browser.smartwindow.customEndpoint
- browser.smartwindow.model
Adding my http endpoint (using LAN IP) seems to work, getting around the check in the regular settings.
This should BE front and centre and what is sold, not ads for a 3rd party.
This sounds great, will have to try it out. Interested in the business value they gain from this, as this appears to be free.
I guess for mistral it's great marketing material.
Edit: Quite interested in the differences of local optimizations... that does allow for specific sensorship, but alas that's already possible in a more direct, reliable way.
I think that beyond marketing this is also a way to collect "AI browsing" trajectories. Not an expert but I think this is the kind of great quality data that can be useful when training LLMs.
These are two companies that seem to be slipping in ethics recently. I'd prefer this over a chrome and openai offering, but I no longer trust these two to necessarily do the right thing regarding privacy
> These are two companies that seem to be slipping in ethics recently
What makes you say that?
This comment thread is one example and not out of recent character
I get that for Mozilla. But what has Mistral done?
They misspelled their own company name:
"...working together to bring Mistal’s scientific innovations to consumers around the world"
A quick proof-read wouldn't have gone amiss here.
I also fail to see how privacy is enhanced by sending private browsing data to a cloud service which then forwards it to an LLM, in addition to storing 'memories' and dredging through my browser history.
On the bright side, at least we know it was typed by human hands!
Perhaps it was planted (satire)
"Private"
Guess we just using buzzwords for marketing and have no morality or principles anymore. Bad timeline and people you should just not trust.
Why would it not be private? Yes data is sent to the servers but the trust that you must have that they don't keep the data is not different from the trust you must have in your password manager or encrypted email or backup storage provider.
All previous similar issues got hand waved with "Mozilla has to make money somehow". I am afraid one day remaining users will decide that Chromium based browsers provide a better experience in most use cases, and now have same privacy guarantees as Firefox. Looking at market share maybe that had already happened and pragmatism won over ideology, and we will all suffer the consequences of such monoculture.
[flagged]
These comments are so annoying and they pop up on literally any news Mozilla releases. Like… ever.
No, Firefox is not malware because it includes an OPTIONAL feature. Yes, chrome is 100x worse than Firefox when it comes to data privacy. Yes, Firefox works in 99.9999% of websites. Yes, the performance of Firefox is about the same as chrome.
Nobody who seriously gives even a single fuck about privacy is going to be dropping Firefox because of this. Because that’s stupid.
The only conclusion, then, is that people who are dropping Firefox because of this don’t care about privacy at all, and just want an excuse to use chrome. But I guess saying “well I like chrome” is too much, so we just lie instead.
It’s fine to use chrome. It IS NOT fine to lie through your teeth and claim you use chrome because Firefox is just as bad for privacy.
Because it’s not, and not a single person would drop Firefox for that reason. You just want to use chrome. So do that, and stop shitting on Firefox as if Mozilla implementing 0.01% of the malware chrome has is the end of the world.
Glad to use ~~Unmozilled Firefox~~ Librewolf
Waterfox also works well!
I remember using Waterfox in the past, I think because they were keeping an old UI. Is that the draw over Librewolf?
For me the draw is that Waterfox has partnerships in order to provide funding so if/when Mozilla truly loses the thread (which some may argue they have) WF can keep going on its own, whereas the Librewolf devs have stated that if push comes to shove they can't maintain an entire browser themselves.
waterfox is downstream of firefox and can't meaningfully match mozilla's maintenance of the core browser
librewolf devs are just more honest about it
Has WF said as much?
Palemoon doesn't embed clankers as far as I know.
SeaMonkey exists. But, I'm sure Mozilla will fuck it up if they ever remember it exists.
Mozilla lost the tech race and with it a lot of users. Firing a lot of their talented developers didn't help. Now Mozilla is giving up the only reason why they kept a core audience of privacy focused users by sending our browsing data to third party servers for inference just at the time when small local models are getting good enough for many/most use cases like summarization or translation. Chrome provides an API for local LLMs. Apple is trying to run as much interference locally as possible. Mozilla is worse than that. Claiming private browsing while sending data to third parties.
I don't care what their data retention policies are, I don't want my data to be sent to others full stop.
Private means it's mine, it's under my control. Handing it to third parties is not private.
I don't understand this negativity. It seems to be a fully optional separate service. No one is losing anything.
What is "AI browsing"?
AI are excellent at ingesting tons of data and getting something relevant from what you asked. I ask it about linux behavior and it will dig through hundreds of patches & mails and will return me informations sourced with links to the patches, mails containing it's affirmations.
I’m fine with cloud features, but I’d like to know what stays on my device and what gets sent out. “Private” is a bit vague.
Didn’t OpenAI tried to build even a new AI browser and failed? I honestly don’t need AI in my browser and would not let an agent to access my login credentials/cookies/saved payment methods for various sites.
It's pretty nice for tedious tasks. I tried to have it look through job listings based on my criteria. It burned through my 5h Claude quota, but what I got was pretty impressive.
I agree on the privacy concerns but this could be useful with a local model, I think.
For me it’s not only privacy, I just don’t trust an agent to use my credentials without doing any destructive task because unlike an MCP, I cannot enforce any permission model on the browser
Not sure if it's a locally run model or in the cloud. If it's the former they don't explain the size on disk and more importantly on memory. I know that google is using local models on the pixel phones, so it's doable. If it's the later, the "Mozilla don't keep the data" is hypocritical and I'm perplexed by the cost associated.
In any case as a french, I prefer my personnal data be exploited by a french company :D
Edit: after reading the comments, it is not a local AI so "Mozilla don't keeep the data and don't train on them" is really hypocritical if they cannot confirm that it's the same on mistral side. And there is still the question of the cost.
I really do not get the obsession with translation, particularly bad translations and translating by default. Most people who browse websites in languages other than the official local one are fluent in that other language.
Google Chrome keeps insisting that Spanish is Galician. PowerPoint insists on changing the spell checker back to the wrong language and ignores setting the language for the whole presentation. On Linux, ChatGPT insists it must enable the spell checker in the local system language, so everything is always red.
Privacy aside, a lot of software seems to be built assuming people cannot be multilingual.
Both are true, a lot of people are multilingual but a lot are also not.
I know a lot of foreigner for which the feature is really useful (local department store, booking a doctor appointment etc.)
I don't know many people, but I do know someone that relies on machine translation to browse the English web. Remember "most" is not "all." The web should be accessible to every single person.
Related, but hardly activity:
"Firefox Smart Window" 21.apr.2026 https://news.ycombinator.com/item?id=47852016
"Firefox Smart Window" 18.aug.2026 https://news.ycombinator.com/item?id=49345179
"Firefox and Exa Partnership" 18.aug.2026 https://news.ycombinator.com/item?id=49346066
this is chatgpt, web gui from mistral built-in Firefox, not for tech savvy users. I think anyone that will complain about this knows very well that you can just give your harness control over firefox through puppeteer, which is much more powerful. could firefox itself become a harness and speak openai api responses? maybe yeah, but why?
Firefox/Mozilla really lost the plot.
Remember when it was the fastest thing around and simple?
Now it's an also ran. AI doesn't help here. Focus on what made it great in the beginning, make it the most performant browser around. Put Fable on the case if need be. Do the browser thing well!
I cannot in a million years imagine using these features. Just be more organised.
I understand why they built this. They probably should and this is probably the right way to do it.
And also yes, I’ve tried Dia and OpenAI’s thing and I can’t for the life of me figure out what to do with the feature.
And then it dawns on me what I really want is a modernized bookmarks system…
[flagged]
Pretty keen on the private, multilingual aspect here. Could be a real game-changer for digging through non-English dev docs.
"Mozilla" and "private" are two phrases that don't go well together, these days.
Hold the AI, just give me a zen/Vivaldi like autohiding vertical tab stack and we're good.
No amount of pixel art will convince me to give up on the freedom to choose.
Neat, how can I turn it off?
Try this: https://support.mozilla.org/en-US/kb/smart-window#w_turn-off...
[flagged]
Cloud inferrance is not private: These providers, even if they don't get hacked, have bribed employees, or outright lie, cannot and will not resist a subpoena or similar.
Mozilla is acting unethically by obfuscating what's on offer here and overstating the privacy and security properties that can be achieved.
It's doubly unfortunate because simple translations are completely reasonable to do with cpu inference.
Late thought: consider what telegram went through over just end to end encryption...
And practically every country has lower privacy standards for data that crosses borders.
Another element of these "claimed to be private but can't keep its promises" is how caustic it is to the trust we have for our friends, family, and business associates.
People here can be privacy aware and well informed and avoid these data slurps-- perhaps go dig up the hidden settings to completely disable it or add firewalls so an errant keypress won't upload your browsing history. That's good.
But anything I share with another person or put on some webpage for another-- perhaps highly trusted person like a doctor or lawyer-- is exposed to them uploading it perhaps completely unwittingly (e.g. they fell for the exaggerated privacy claims) or due to an innocent misclick.
Normalizing privacy fails like this undermines the ability of even the most informed and aware people to opt-out.
So now mistral has partnership with browser + user logs from Mozilla. Does it means that they are trying to build their own search index? Because it’s kinda huge inference investment in exchange for what? Any other ideas why they need this?
Nobody asked. Give me back my RAM.
I remember the days when people justified Google Chrome's maximalist RAM usage by saying it was a good thing because it was allocating what the program actually needed to run quickly and efficiently, and it was better to grab what it actually needed then to operate inefficiently while using fewer system resources.
For a while that was even the majority view, but it never quite made sense to me.
What the fuck is the point of this?
What is Mistral bringing to the table for Firefox? There is nothing "open" about this in any way.
FF needs to either profile itself as the no-ai-by-default browser, or it needs to just have OpenAI/Anthropic/Mistral/DeepSeek bid for the default spot - like they do with Google search.
I'm happy Mistral exists. I'm happy Firefox exists.
None of this shows any synergy i'm excited about.
Maybe FF believes there is a group of users who are still on the fence about using FF - until they can pitch them a first-class build-in AI story that goes with the anti-establishment vibe?
I somehow doubt that's the pitch & potential userbase they should be focussing on.
I certainly hope one of Ironfox or Librewolf will strip this shit out by default.
>>Mistral has been releasing open weight, frontier models since our first release.
yeah ?
[flagged]
[dead]
[dead]
[dead]
[dead]
[dead]
[dead]
[flagged]
Please don't.
This will put duo lingus out of business.