The license seems really restrictive for something intended to be built on. Entire countries are forbidden to use it at all (European Union, the United Kingdom and South Korea), and you are responsible for making sure no one in those places uses your derivative.
That's more often than not the point - they get to pick and choose where to enforce, likely based on whether or not said user is in competition with them.
Brilliant! Base model releases are exactly what we need right now. This is the reliable way to do alignment. Releasing the base model then letting everyone on the planet do whatever they want with it is probably what ends up preventing extreme value drift through the intelligence explosion. Millions of people would try to align these base models in their own way, and we'll probably all fail, but those models would drift in different directions so their actions cancel eachother out in the real world. The end result is alignment to humanity's average objective function, at the civilization level. It's a messy but very surefire way to align.
There will still be a different kind of value drift from increasing intelligence itself, but that's unavoidable no matter what we do. Releasing open models slows that down too because it discourages investment in future training runs. And this strategy works unilaterally without cooperation - every open model release slows it down by a little bit. This is the actual way to pace the frontier unlike the regulatory capture stunts Anthropic and OpenAI are attempting while pontificating about their moral superiority.
So, what exactly can a base model do? Is this like the pre-reasoning models that just autocompleted, without any alignment, reasoning, etc optimizations?
A large base model is ~90% of the effort/cost to making a frontier model, and by itself is not useful for much. There aren't many controversial decisions to make when training a base model - you mostly just train a model as big as possible on as much human-generated data as possible.
But it makes it far easier for other parties to do that last 10% to train it to be useful.
However, there are many choices one makes in the last 10% of training which will impact the models behaviour when it comes to AI safety, political opinions, willingness to help make a bomb, etc etc.
> A large base model is ~90% of the effort/cost to making a frontier model
This kind of figures is often quoted but where does that estimation comes from? Isn't that legacy number from the era when all the post training WS just instruction following and RLHF? I'd be surprised if current generation if reinforcement learning was still a few percent of the training cost.
AIUI more recent estimates have about half of the total training compute, or more, being used for post-training/fine-tuning. Post training a base model (even for simple chat, let alone thinking/reasoning or performance on verifiable tasks) requires reinforcement learning, which is quite hard.
Trump just realised that he can't use MAGA for Making SI Great Again, probably giving an executive order to reverse SI to AI again, to be able to claim Make AI Great Again
I am PLEASED TO ANNOUNCE that SI will be America’s 51st STATE and will be RENAMED to “American Intelligence”! Thank you for your attention to this matter!
...and when it makes a mistake, it blames anything or anyone else but itself, or insists that what it did is perfect the way it is and that it's still the smartest model in the world?
Anyone suggest a framework to further train this model on a multi-node cluster with infiniband? NeMo Framework? "Plain" HF TRL+Accelerate? Opinionated suggestions are welcome.
This is awesome. ERNIE seems more powerful than Qwen3.8-27B and Deepseek-v4-Flash, which AFAIK are currently the best base models... it'll be very interesting to experiment with when it hits OpenCode. Base models are also excellent for businesses who run models locally, as they can be trained in various house rules and house styles.
They're comparing it to Deepseek-v4-Flash, not 4.1 and not the 4.1 1017 update, so essentially two versions behind the latest version of Deepseek Flash.
Am I the only one to feel sad that the name means "Enhanced Representation through kNowledge IntEgration", rather than being a homage to a very famous 1950s random number generator?
It's a homage to BERT (Bidirectional Encoder Representations from Transformers), Google's influential 2018 Transformer model that wasn't autoregressive. https://arxiv.org/abs/1810.04805v2
A chunk better than Deepseek 1017, at half the parameter count, BUT it's only a base model, not finetuned to follow instructions.
Strange to release a model in this state. OpenAI is also "probably" releasing new models (sources within OpenAI talking about Sol 6.2 and the like). We must be getting really close to the Qwen4 releases.
Many others, e.g. Google and NVIDIA, have also released such BASE models.
For those with the knowledge and resources needed to finalize the training, BASE models are much more useful than completely trained models.
BASE models are the next best thing after truly open-source LLMs, where you would have access to the complete dataset used for training. Besides the fact that nobody who had used a really big dataset wants to show it, due to possible legal problems, having access to the original dataset would not be helpful for most people, because they would not be able to use it to make an LLM, due to its huge size and due to the vast resources required for pre-training.
On the other hand, continuing training from a BASE model, which is greatly compressed in comparison with the original dataset, is feasible for many.
The license seems really restrictive for something intended to be built on. Entire countries are forbidden to use it at all (European Union, the United Kingdom and South Korea), and you are responsible for making sure no one in those places uses your derivative.
That seems... impossible to enforce.
That's more often than not the point - they get to pick and choose where to enforce, likely based on whether or not said user is in competition with them.
Brilliant! Base model releases are exactly what we need right now. This is the reliable way to do alignment. Releasing the base model then letting everyone on the planet do whatever they want with it is probably what ends up preventing extreme value drift through the intelligence explosion. Millions of people would try to align these base models in their own way, and we'll probably all fail, but those models would drift in different directions so their actions cancel eachother out in the real world. The end result is alignment to humanity's average objective function, at the civilization level. It's a messy but very surefire way to align.
There will still be a different kind of value drift from increasing intelligence itself, but that's unavoidable no matter what we do. Releasing open models slows that down too because it discourages investment in future training runs. And this strategy works unilaterally without cooperation - every open model release slows it down by a little bit. This is the actual way to pace the frontier unlike the regulatory capture stunts Anthropic and OpenAI are attempting while pontificating about their moral superiority.
So, what exactly can a base model do? Is this like the pre-reasoning models that just autocompleted, without any alignment, reasoning, etc optimizations?
A large base model is ~90% of the effort/cost to making a frontier model, and by itself is not useful for much. There aren't many controversial decisions to make when training a base model - you mostly just train a model as big as possible on as much human-generated data as possible.
But it makes it far easier for other parties to do that last 10% to train it to be useful.
However, there are many choices one makes in the last 10% of training which will impact the models behaviour when it comes to AI safety, political opinions, willingness to help make a bomb, etc etc.
> A large base model is ~90% of the effort/cost to making a frontier model
This kind of figures is often quoted but where does that estimation comes from? Isn't that legacy number from the era when all the post training WS just instruction following and RLHF? I'd be surprised if current generation if reinforcement learning was still a few percent of the training cost.
AIUI more recent estimates have about half of the total training compute, or more, being used for post-training/fine-tuning. Post training a base model (even for simple chat, let alone thinking/reasoning or performance on verifiable tasks) requires reinforcement learning, which is quite hard.
A base model is what you get just from next token prediction training. Think GPT-2/3 but bigger.
Withtout RLHF, instruction-following training, or RLVR to develop reasoning.
I am not sure to follow, why couldn't the vast majority of uses of this base model be nefarious/selfish?
Isn't it like saying "let's give infinite power to all humans on the planet, nothing can go wrong since they define what is humanity" ?
ERNIE feels old, like BERT. They either need a rebrand or fully embrace retro-dawn-of-ML vibes
BERT models are still amazing for what they do. Even the latest JEV/KEV/SEV/BEV/REV/TREV models can't beat a well-trained BERT.
do you have examples of open weight BERT models from hugging face which are trained for specific use cases?
Very old .. https://en.wikipedia.org/wiki/Premium_Bonds#ERNIE
Bert and Ernie. Ring a bell? Sesame Street
I'd try DONALD for sure, if I wanted to kill a project.
Make AI Great Again?
SI*
Curious he didn't rename it American Intelligence instead.
Yes?
Lake Intelligence
Trump just realised that he can't use MAGA for Making SI Great Again, probably giving an executive order to reverse SI to AI again, to be able to claim Make AI Great Again
I am PLEASED TO ANNOUNCE that SI will be America’s 51st STATE and will be RENAMED to “American Intelligence”! Thank you for your attention to this matter!
But they didn't give a DONALD.
> I'd try DONALD for sure, if I wanted to kill a project.
The current version is threatening to veto all my prompts.
...and when it makes a mistake, it blames anything or anyone else but itself, or insists that what it did is perfect the way it is and that it's still the smartest model in the world?
Anyone suggest a framework to further train this model on a multi-node cluster with infiniband? NeMo Framework? "Plain" HF TRL+Accelerate? Opinionated suggestions are welcome.
This is awesome. ERNIE seems more powerful than Qwen3.8-27B and Deepseek-v4-Flash, which AFAIK are currently the best base models... it'll be very interesting to experiment with when it hits OpenCode. Base models are also excellent for businesses who run models locally, as they can be trained in various house rules and house styles.
They're comparing it to Deepseek-v4-Flash, not 4.1 and not the 4.1 1017 update, so essentially two versions behind the latest version of Deepseek Flash.
DeepSeek 4.1 Flash is twice the size of this and difficult to run locally. Makes more sense to compare it to DS4F.
Qwen3.8-27B has no base model version in the wild that I can see?
Isn’t the latest Qwen 3 public base model… Qwen 3?
Despite the naming Qwen3 and Qwen3.5+ have very little in common in terms of architecture. (And neither does 3.8-Flash-next …)
Absolutely glad to see the release of a base model, I think the trend away from that has been an issue for the field.
That's an interesting size, it will fit into 2x128GB at decent (≥4bpw) quantization.
Am I the only one to feel sad that the name means "Enhanced Representation through kNowledge IntEgration", rather than being a homage to a very famous 1950s random number generator?
https://en.wikipedia.org/wiki/Premium_Bonds#ERNIE
For me I see ERNIE as BERT's long time friend and room mate.
How do I look Ernie? ... with your eyes, Bert.
It's a homage to BERT (Bidirectional Encoder Representations from Transformers), Google's influential 2018 Transformer model that wasn't autoregressive. https://arxiv.org/abs/1810.04805v2
anyone tried this? is it good?
It's not instruction finetuned, so no, it's not good to use for anything like coding agents.
chinnese model slop
OT
dropped is the worst word in this context because it can both mean ‘released’ or ‘terminated/ended’
Baidu dropped ERNIE 5.5 Flash Base = Baidu released a new version of their software
Baidu dropped ERNIE 5.5 Flash Base = Baidu ended the support of their software
I think in contemporary parlance this sufficiently clear. Nobody thought massdrop was where you went to buy end-of-life products.
TLDR: Ernie 5.5 Flash base 280B params 14B active
A chunk better than Deepseek 1017, at half the parameter count, BUT it's only a base model, not finetuned to follow instructions.
Strange to release a model in this state. OpenAI is also "probably" releasing new models (sources within OpenAI talking about Sol 6.2 and the like). We must be getting really close to the Qwen4 releases.
Many others, e.g. Google and NVIDIA, have also released such BASE models.
For those with the knowledge and resources needed to finalize the training, BASE models are much more useful than completely trained models.
BASE models are the next best thing after truly open-source LLMs, where you would have access to the complete dataset used for training. Besides the fact that nobody who had used a really big dataset wants to show it, due to possible legal problems, having access to the original dataset would not be helpful for most people, because they would not be able to use it to make an LLM, due to its huge size and due to the vast resources required for pre-training.
On the other hand, continuing training from a BASE model, which is greatly compressed in comparison with the original dataset, is feasible for many.
I believe they compare to DSv4, not DSv4.1 1017.
Specifically V4-Flash-Base from April.
> Strange to release a model in this state.
No, they explain why they do it.