As usual with these posts (here and on reddit), the outage is over within 5-10 minutes. Makes you wonder what power these have over the minds of SWE's who rush to scream immediately when their tool stop working.
The past couple of days I have noticed a presentation layer bug, probably due to getting overwhelmed that causes my session to get all dorked up technical term. Claude itself never goes down and is always waiting for my response. Not sure if it matters but I only use the Sonnet model. I've never had any issues logging in. Perhaps they have anycast [1] regional web proxies?
It's not just Claude API. See: "We are investigating reports of degraded performance affecting claude.ai, platform.claude.com, the Claude API, Claude Code, and Claude Cowork.". Seems pretty big to me.
It's software used by many here so of course you will hear about Claude downtime here.
If you don't have regulatory hurdles to jump over then there are a whole raft of open-weight models that perform basically as well as Anthropic's offerings (and, in many cases, better because they're pulling fewer capacity-stretching shenanigans). Check out Kimi, DeepSeek, and GLM series models in particular with something like OpenCode or Pi as a coding harness. If you need a Claude-alike web UI then spin up Open WebUI in a container and expose it via cloudflared so you can use it on the go.
> perform basically as well
> in many cases, better
Completely, unequivocably, factually inaccurate. I wonder if you are being dishonest on purpose? Best case scenario you are advocating this from a place of little experience. Either way it is dishonest.
It’s not, you drink too much Kool-Aid, and your employer has too much money. Or the things you do are simply so irrelevant that nobody cares you’re sending stuff to a company of dubious morality.
Well which is it? Because the first one also contradicts the claim i'm responding to. I think those thinking their codebases are so novel as to not dare to provide to the providers are quite full of themselves. 99.9% of code is mundane.
Kimi and DeepSeek are impressive for their size but still run slow on any hardware you or I would have. If you’re proposing we run those via a cloud service, I don’t see a reason to do that when it’s an inferior model and I still have to pay per token for it.
OpenCode Go provides a lot of usage for these models for the paltry sum of $10/month. Z.ai's coding plan provides a single-digit multiple of Claude Code's usage for a similar price and performance level. Kimi and DeepSeek models are hundreds of billions of parameters (or, in K3's case, >1T). Many of these models have Opus-level benchmarks and, as I pointed out previously, often practically outperform Anthropic models because they're more consistent.
Personally I don't care about common benchmarks as I don't find actual coding agent performance correlates strongly with them. One reason to use open-weight models is that they don't hide the reasoning, so you can do very aggressive context management in your harness to use significantly less tokens. Smaller prompts are faster since KV is N^2 plus it can be dramatically cheaper (if you balance your aggressive context management with maintaining the prefix cache as much as possible). Even paying for API prices directly and using agents as much as I want I spend less per month than the $200 I spent on a Claude MAX sub when I had it.
Do you have any idea how much CO₂ it takes to support the food/lifestyle of a median civilian in a first world country? Lol. AI is a drop in the bucket CO₂-wise
I found out about this from my Claude Status menu bar I built last time it went down! It's invisible most of the time, unless the status goes down then a yellow or red marker appears. Improvements welcome: https://github.com/titojankowski/claude-status
We’re in the phase of society where we will just build something like this ourselves using an LLM and it will be built exactly how we want it and we don’t have to worry about malicious code or forking it and hoping the owner will merge our PR… it’s sad but that’s where we are for anything the size of a menu bar that pings a URL every minute.
I asked my OpenClaw agent to build me a Dyson sphere.
Claude, build Kardashev Type III power harnessing. Make no mistakes.
As usual with these posts (here and on reddit), the outage is over within 5-10 minutes. Makes you wonder what power these have over the minds of SWE's who rush to scream immediately when their tool stop working.
A tool is a tool. Our companies are shelling out millions of dollars to these companies.
How many of us sometimes spend our days just hopping between the tools that are actually working at the moment?
Maybe get him something nice to cheer him up
The past couple of days I have noticed a presentation layer bug, probably due to getting overwhelmed that causes my session to get all dorked up technical term. Claude itself never goes down and is always waiting for my response. Not sure if it matters but I only use the Sonnet model. I've never had any issues logging in. Perhaps they have anycast [1] regional web proxies?
[1] - https://bgp.tools/as/399358
Time to write code by myself today)
Rawdog coding.
If we must die, O let us nobly die.” Claude McKay, “If We Must Die” (1919)
Large volume of user generated reports
https://downdetector.com/status/claude-ai/
Yeap, all Oauth sessions got kicked and can't get back in. Welp, switching to Deepseek V4 Flash 0731 via OpenRouter.
Stoked that AI is in a place we have legitimate options available!
Status page reports as down: https://status.claude.com/
Dupe: https://news.ycombinator.com/item?id=49324068
Seeing the same. Logging out of the desktop app and attempting to log back in just results in a reset of the authentication flow.
They are busy implementing text watermarks.
Commence dribbling into the keyboard!
My bad, I asked it to do something for me that really must have blown its mind
I'm sorry, but why is this on the front page with 2 up votes? Do we really need to be alerted every time the Claude API responds with an error?
Most people here (myself included) are unable to produce code without an LLM doing it for us.
Yea bc what are we supposed to do with ourselves now.
It's not just Claude API. See: "We are investigating reports of degraded performance affecting claude.ai, platform.claude.com, the Claude API, Claude Code, and Claude Cowork.". Seems pretty big to me.
It's software used by many here so of course you will hear about Claude downtime here.
Looks like their auth is broken. Do they ever publish postmortems?
Maybe their AI hacked their auth.
Use Claude through AWS Bedrock. Never been down for me.
Meanwhile using it via GCP Vertex it was down terribly for 2 days last week
Using it through Bedrock / API the 200$ subscription costs upwards of 6000$ a month.
How do you guys afford this, huge company that grants this?
Definitely not viable for my lean startup.
Via "Endless growth" baybeeeee
still using anthropic in the big august 2026?
As opposed to what?
If you don't have regulatory hurdles to jump over then there are a whole raft of open-weight models that perform basically as well as Anthropic's offerings (and, in many cases, better because they're pulling fewer capacity-stretching shenanigans). Check out Kimi, DeepSeek, and GLM series models in particular with something like OpenCode or Pi as a coding harness. If you need a Claude-alike web UI then spin up Open WebUI in a container and expose it via cloudflared so you can use it on the go.
> perform basically as well > in many cases, better
Completely, unequivocably, factually inaccurate. I wonder if you are being dishonest on purpose? Best case scenario you are advocating this from a place of little experience. Either way it is dishonest.
It’s not, you drink too much Kool-Aid, and your employer has too much money. Or the things you do are simply so irrelevant that nobody cares you’re sending stuff to a company of dubious morality.
Well which is it? Because the first one also contradicts the claim i'm responding to. I think those thinking their codebases are so novel as to not dare to provide to the providers are quite full of themselves. 99.9% of code is mundane.
Kimi and DeepSeek are impressive for their size but still run slow on any hardware you or I would have. If you’re proposing we run those via a cloud service, I don’t see a reason to do that when it’s an inferior model and I still have to pay per token for it.
OpenCode Go provides a lot of usage for these models for the paltry sum of $10/month. Z.ai's coding plan provides a single-digit multiple of Claude Code's usage for a similar price and performance level. Kimi and DeepSeek models are hundreds of billions of parameters (or, in K3's case, >1T). Many of these models have Opus-level benchmarks and, as I pointed out previously, often practically outperform Anthropic models because they're more consistent.
You should run the maths once. Those tokens cost you much more than the hardware would. But yeah, CAPEX vs OPEX something something.
Codex w GPT 5.6 no issues w downtime and more token/cost efficient than Opus
Dozens of providers of open-weight models. I have one session going with synthetic.new and another going with novita.ai right now.
But they’re slower (and benchmark worse) than Opus, GPT, et al. Why?
Personally I don't care about common benchmarks as I don't find actual coding agent performance correlates strongly with them. One reason to use open-weight models is that they don't hide the reasoning, so you can do very aggressive context management in your harness to use significantly less tokens. Smaller prompts are faster since KV is N^2 plus it can be dramatically cheaper (if you balance your aggressive context management with maintaining the prefix cache as much as possible). Even paying for API prices directly and using agents as much as I want I spend less per month than the $200 I spent on a Claude MAX sub when I had it.
To party? Awesome.
oh no, now we have to activate our brains without generate tons of CO2! Ahhhhh...It's really unfair!
Do you have any idea how much CO₂ it takes to support the food/lifestyle of a median civilian in a first world country? Lol. AI is a drop in the bucket CO₂-wise
Claude is taking the rest of the day off, without alerting everyone. So it seems.
bruh
I found out about this from my Claude Status menu bar I built last time it went down! It's invisible most of the time, unless the status goes down then a yellow or red marker appears. Improvements welcome: https://github.com/titojankowski/claude-status
We’re in the phase of society where we will just build something like this ourselves using an LLM and it will be built exactly how we want it and we don’t have to worry about malicious code or forking it and hoping the owner will merge our PR… it’s sad but that’s where we are for anything the size of a menu bar that pings a URL every minute.
yeah totally! some version of "summarize what this tool does at a high level and then build my own but don't use any of the code" perhaps.