I don’t do ‘evals’, but I do process billions of tokens every month, and I’ve found these small Nvidia models to be the best by far for their size currently.
As someone else mentioned, the GPT-OSS models are also quite good (though I haven’t found how to make them great yet, though I think they might age well like the Llama 3 models did and get better with time!).
But for a defined task, I’ve found task compliance, understanding, and tool call success rates to be some of the highest on these Nvidia models.
For example, I have a continuous job that evaluates if the data for a startup company on aVenture.vc could have overlapping/conflated two similar but unrelated companies for news articles, research details, investment rounds, etc… which is a token hungry ETL task! And I recently retested this workflow on the top 15 or so models today with <125b parameters, and the Nvidia models were among the best performing for this type of work, particularly around non-hallucination if given adequate grounding.
Also, re: cost - I run local inference on several machines that run continuously, in addition to routing through OpenRouter and the frontier providers, and was pleasantly surprised to find that if I’m a paying customer of OpenRouter otherwise, the free variant there from Nvidia is quite generous for limits, too.
You may want to use the new "derestricted" variants of gpt-oss. While the ostensible goal of these variants is to de-censor them, it ends up removing the models' obsession with policy and wasting thinking tokens that could be used towards actually reasoning through a problem.
I recently pitted gpt-oss 120b against Qwen3-Next 80b on a lot of internal benchmarks (for production use), and for me, gpt-oss was slightly slower (vLLM, both fit in VRAM), much worse at multilingual tasks (33 languages evaluated), and had worse instruction following (e.g., Qwen3-Next was able to reuse the same prompts I used for Gemma3 perfectly, while gpt-oss struggled and RAG benchmarks suddenly went from 90% to 60% without additional prompt engineering).
And that's with Qwen3-Next being a random unofficial 4-bit quant (compared to gpt-oss having native support) + I had to disable multi-token prediction in Qwen3-Next because vLLM crashed with it.
Has someone here tried both gpt-oss 120b and Qwen3-Next 80b? Maybe I was doing something wrong because I've seen a lot of people praise gpt-oss.
Completely agree. I was working on something with TensorRT LLM and threw Nemotron in there more on a whim. It completely mopped the floor with other models for my task (text style transfer), following joint moderation with another LLM & humans. Really impressed.
He's just saying anecdotally these models are good. A reasonable response might be "have you systematically evaluated them?". He has pre-answered - no.
The Nano model isn’t pretrained in FP4, only Super and Ultra are. And posttraining is not in FP4, so the posttrained weights of these models are not native FP4.
I've noticed that open models have made huge efficiency gains in the past several months. Some amount of that is explainable as architectural improvements but it seems quite obvious that a huge portion of the gains come from the heavy use of synthetic training data.
In this case roughly 33% of the training tokens are synthetically generated by a mix of other open weight models. I wonder if this trend is sustainable or if it might lead to model collapse as some have predicted. I suspect that the proliferation of synthetic data throughout open weight models has lead to a lot of the ChatGPT writing style replication (many bullet points, em dashes, it's not X but actually Y, etc).
Im upvoting, im happy to finally see open source model with commercial use from Nvidia as most of the models ive been checking from you guys couldnt be used in commercial settings. Bravo Nvidia!
I would say it is weird, that NVidia competes with own customers but looking back at "Founders Edition" cards maybe it isn't that weird at all. The better question probably is - with every big corporation having its own LLM, what exactly is OpenAI moat that would explain their valuation?
They both believe the product people focus on will commoditize. Tesla realized early that EVs without autonomy are a dead end for long-term dominance, just as NVIDIA believes models without infrastructure are a dead end for durable AI profits.
If it's intelligence + speed you want, nothing comes close to GPT-OSS-120B on Cerebras or Groq.
However, this looks like it has great potential for cost-effectiveness. As of today it's free to use over API on OpenRouter, so a bit unclear what it'll cost when it's not free, but free is free!
That's unlikely. Cerebras doesn't speed up everything. Can it speed up everything? I don't know, I'm not an insider. But does it speed up everything? That is evidently not the case. Their page [1] lists only 4 production models and 2 preview models.
I find it really interesting that it uses a Mamba hybrid with Transformers. Is it the only significant model right now using (at least partially) SSM layers? This must contribute to lower VRAM requirements right? Does it impact how KV caching works?
Is it fair to view this release as Nvidia strategically flexing that they can compete with their own customers in the model layer -- that they can be as vertically integrated as, say, GDM?
can it understand input in and generate output for different language tokens? does it know narrow IPA transcription of sentences in arbitrary languages?
I used Nemotron 3 nana on LM Studio yesterday on my 32G M2-Pro mac mini. It is fast and passed all of my personal tool use tests, and did a good job analyzing code. Love it.
Today I ran a few simple cases on Ollama, but not much real testing.
Kind of depends on your mac, but if it's a relatively recent apple silicon model… maybe, probably?
> Nemotron 3 Nano is a 3.2B active (3.6B with embeddings) 31.6B total parameter model.
So I don't know the exact math once you have a MoE, but 3.2b will run on most anything, 31.6b and you're looking at needing a pretty large amount of ram.
Given Mac bandwidth, you'll generally want to load the whole thing in RAM. You get speed benefits based on smaller-size active experts, since the Mac compute is slow compared to Nvidia hardware. This should be relatively snappy on a Mac, if you can load the entire thing.
The claim that a small, fast, and decently accurate model makes a good foundation for agentic workloads seems like a reasonable claim.
However, is cost the biggest limiting factor for agent adoption at this point? I would suspect that the much harder part is just creating an agent that yields meaningful results.
No, I really don't think cost is the limiting factor- it's tooling and competent workforce to implement it. Every company of any substantial size, or near enough, is trying to implement and hire for those roles, and the # of people familiar with the specific tooling + lack of maturity in tooling increasing the learning curve, these are the bottlenecks.
This has been my major concern, so much do that I'm going to be launching a tool to handle this specific task: agent conception and testing. There is so little visibility in the tools I've used that debug is just a game of whackamole.
after testing it for a little I am pretty disappointed. While I do get 90 token per second out of it from my M4 Pro which is more than enough for a real world use case, the quality is just not there. I gave it a codebase that it should analyze and answer me some questions and it started hallucinating right away. No replacement for a "real" coding agent - maybe for other agentic work like sorting emails though.
FWIW It runs on my 9060xt(AMD) 16gb, without any tweaks just fine. It's very useable.
I asked it to write a prime sieve in c#, started responding in .38 seconds, and wrote an implementation @ 20 tokens/sec
I don’t do ‘evals’, but I do process billions of tokens every month, and I’ve found these small Nvidia models to be the best by far for their size currently.
As someone else mentioned, the GPT-OSS models are also quite good (though I haven’t found how to make them great yet, though I think they might age well like the Llama 3 models did and get better with time!).
But for a defined task, I’ve found task compliance, understanding, and tool call success rates to be some of the highest on these Nvidia models.
For example, I have a continuous job that evaluates if the data for a startup company on aVenture.vc could have overlapping/conflated two similar but unrelated companies for news articles, research details, investment rounds, etc… which is a token hungry ETL task! And I recently retested this workflow on the top 15 or so models today with <125b parameters, and the Nvidia models were among the best performing for this type of work, particularly around non-hallucination if given adequate grounding.
Also, re: cost - I run local inference on several machines that run continuously, in addition to routing through OpenRouter and the frontier providers, and was pleasantly surprised to find that if I’m a paying customer of OpenRouter otherwise, the free variant there from Nvidia is quite generous for limits, too.
You may want to use the new "derestricted" variants of gpt-oss. While the ostensible goal of these variants is to de-censor them, it ends up removing the models' obsession with policy and wasting thinking tokens that could be used towards actually reasoning through a problem.
>the GPT-OSS models are also quite good
I recently pitted gpt-oss 120b against Qwen3-Next 80b on a lot of internal benchmarks (for production use), and for me, gpt-oss was slightly slower (vLLM, both fit in VRAM), much worse at multilingual tasks (33 languages evaluated), and had worse instruction following (e.g., Qwen3-Next was able to reuse the same prompts I used for Gemma3 perfectly, while gpt-oss struggled and RAG benchmarks suddenly went from 90% to 60% without additional prompt engineering).
And that's with Qwen3-Next being a random unofficial 4-bit quant (compared to gpt-oss having native support) + I had to disable multi-token prediction in Qwen3-Next because vLLM crashed with it.
Has someone here tried both gpt-oss 120b and Qwen3-Next 80b? Maybe I was doing something wrong because I've seen a lot of people praise gpt-oss.
gpt-oss is STEM-maxxed, so I imagine most of the praise comes from people using it for agentic coding.
> We trained the models on a mostly English, text-only dataset, with a focus on STEM, coding, and general knowledge.
https://openai.com/index/introducing-gpt-oss/
Completely agree. I was working on something with TensorRT LLM and threw Nemotron in there more on a whim. It completely mopped the floor with other models for my task (text style transfer), following joint moderation with another LLM & humans. Really impressed.
Would you mind sharing what hardware/card(s) you're using? And is https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B... one of the ones you've tested?
Support for this landed in llama.cpp recently if anyone is interested in running it locally.
What do you mean about not doing evals? Just literally that you don’t run any benchmarks or do you have something against them?
He's just saying anecdotally these models are good. A reasonable response might be "have you systematically evaluated them?". He has pre-answered - no.
Not OP, but perhaps they mean not putting too much faith in common benchmarks (thanks to benchmaxxing).
Very interesting release:
* Hybrid MoE: 2-3x faster than pure MoE transformers
* 1M context length
* Trained on NVFP4
* Open Source! Pretraining, mid-training, SFT and RL dataset released (SFT HF link is 404...)
* Open model training recipe (coming soon)
Really appreciate Nvidia being the most open lab but they really should make sure all the links/data are available on day 0.
Also interesting that the model is trained in NVFP4 but the inference weights are FP8.
The Nano model isn’t pretrained in FP4, only Super and Ultra are. And posttraining is not in FP4, so the posttrained weights of these models are not native FP4.
I love how detailed and transparent the data set statistics are on the huggingface pages. https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B...
I've noticed that open models have made huge efficiency gains in the past several months. Some amount of that is explainable as architectural improvements but it seems quite obvious that a huge portion of the gains come from the heavy use of synthetic training data.
In this case roughly 33% of the training tokens are synthetically generated by a mix of other open weight models. I wonder if this trend is sustainable or if it might lead to model collapse as some have predicted. I suspect that the proliferation of synthetic data throughout open weight models has lead to a lot of the ChatGPT writing style replication (many bullet points, em dashes, it's not X but actually Y, etc).
We just evaluated Nemotron-3 for Vectara's hallucination leaderboard.
It scores at 9.6% hallucination rate, similar to qwen3-next-80b-a3b-thinking (9.3%) but of course it is much smaller.
https://github.com/vectara/hallucination-leaderboard
Im upvoting, im happy to finally see open source model with commercial use from Nvidia as most of the models ive been checking from you guys couldnt be used in commercial settings. Bravo Nvidia!
Just wondering is any commercial restriction can be considered open source at all? Even the most stringent GPL allows you to commercialize [1].
But we are talking about LLM model here not software, but the same principle should applies.
[1] Open-source license:
https://en.wikipedia.org/wiki/Open-source_license
I would say it is weird, that NVidia competes with own customers but looking back at "Founders Edition" cards maybe it isn't that weird at all. The better question probably is - with every big corporation having its own LLM, what exactly is OpenAI moat that would explain their valuation?
I wonder if they also want to create more of a market for their products such as the DGX Spark.
They and Tesla know something no one else does.
Can you tell us more? I’m curious to hear what is behind this implication.
A guess:
They both believe the product people focus on will commoditize. Tesla realized early that EVs without autonomy are a dead end for long-term dominance, just as NVIDIA believes models without infrastructure are a dead end for durable AI profits.
(Am I close?)
If it's intelligence + speed you want, nothing comes close to GPT-OSS-120B on Cerebras or Groq.
However, this looks like it has great potential for cost-effectiveness. As of today it's free to use over API on OpenRouter, so a bit unclear what it'll cost when it's not free, but free is free!
https://openrouter.ai/nvidia/nemotron-3-nano-30b-a3b:free
> nothing comes close to GPT-OSS-120B on Cerebras
That's temporary. Cerebras speeds up everything, so if Nemotron is good quality, it's just a matter of time until they add it.
That's unlikely. Cerebras doesn't speed up everything. Can it speed up everything? I don't know, I'm not an insider. But does it speed up everything? That is evidently not the case. Their page [1] lists only 4 production models and 2 preview models.
[1] https://inference-docs.cerebras.ai/models/overview
They need to compile the model for their chips. Standard transformers are easier, so GPT-OSS, Qwen, GLM, etc if there is demand, they will deploy it.
Nemotron on the other hand is a hybrid (Transformer + Mamba-2) so it will be more challenging to compile it on Cerebras/Groq chips.
(Me thinks Nvidia is purposefully picking architecture+FP4 that is easy to ship on Nvidia chips, but harder for TPU or Cerebras/Groq to deploy)
I find it really interesting that it uses a Mamba hybrid with Transformers. Is it the only significant model right now using (at least partially) SSM layers? This must contribute to lower VRAM requirements right? Does it impact how KV caching works?
Is it fair to view this release as Nvidia strategically flexing that they can compete with their own customers in the model layer -- that they can be as vertically integrated as, say, GDM?
Nemotron now works on LM Studio if you update the runtime (from the settings > Runtime screen).
The default chat template is incorrect though and will fail but I published a corrected one you can replace it with: https://gist.github.com/omarkamali/a594b6cb07347f501babed489...
I was just using the embeddings model last night. Boy is it slow. Nice results but this 5090 isn't cutting it.
I'm guessing there's some sophistication in the instrumentation I'm just not up to date with.
can it understand input in and generate output for different language tokens? does it know narrow IPA transcription of sentences in arbitrary languages?
The article seem to focus on the nano model. Where are the details of the larger ones?
> We are releasing the Nemotron 3 Nano model and technical report. Super and Ultra releases will follow in the coming months.
Any chance of running this nano model on my Mac?
I used Nemotron 3 nana on LM Studio yesterday on my 32G M2-Pro mac mini. It is fast and passed all of my personal tool use tests, and did a good job analyzing code. Love it.
Today I ran a few simple cases on Ollama, but not much real testing.
There's MLX versions of the model, so yes. LM Studio hasn't updated their mlx-lm runtime yet though, you'll get an exception.
But if you're OK running it without a UI wrapper, mlx_lm==0.30.0 will serve you fine.
Looks like LM Studio just updated the MLX runtime, so there's compatibility now.
Yep! 60t/s on the 8 bit MLX on an M4 Pro with 64GB of RAM.
LMStudio and 32+ gb of RAM.
https://lmstudio.ai/models/nemotron-3
Simplest to just install it from the app.
Kind of depends on your mac, but if it's a relatively recent apple silicon model… maybe, probably?
> Nemotron 3 Nano is a 3.2B active (3.6B with embeddings) 31.6B total parameter model.
So I don't know the exact math once you have a MoE, but 3.2b will run on most anything, 31.6b and you're looking at needing a pretty large amount of ram.
Given Mac bandwidth, you'll generally want to load the whole thing in RAM. You get speed benefits based on smaller-size active experts, since the Mac compute is slow compared to Nvidia hardware. This should be relatively snappy on a Mac, if you can load the entire thing.
running it on my M4 @ 90tps, takes 18GB of RAM.
If it uses 18GB of RAM, you're not using the official model (released in BF16 and FP8), but a quantization of unknown quality.
If you write "M4", you mean M4 and not M4 Pro or M4 Max?
M2 Max @ 17tps btw
The claim that a small, fast, and decently accurate model makes a good foundation for agentic workloads seems like a reasonable claim.
However, is cost the biggest limiting factor for agent adoption at this point? I would suspect that the much harder part is just creating an agent that yields meaningful results.
No, I really don't think cost is the limiting factor- it's tooling and competent workforce to implement it. Every company of any substantial size, or near enough, is trying to implement and hire for those roles, and the # of people familiar with the specific tooling + lack of maturity in tooling increasing the learning curve, these are the bottlenecks.
This has been my major concern, so much do that I'm going to be launching a tool to handle this specific task: agent conception and testing. There is so little visibility in the tools I've used that debug is just a game of whackamole.
Did you see this HN submission? https://news.ycombinator.com/item?id=46242838
It seems similar to what you're describing.
I did not. Thanks for the heads up!
after testing it for a little I am pretty disappointed. While I do get 90 token per second out of it from my M4 Pro which is more than enough for a real world use case, the quality is just not there. I gave it a codebase that it should analyze and answer me some questions and it started hallucinating right away. No replacement for a "real" coding agent - maybe for other agentic work like sorting emails though.
Is it just me or is Nvidia trolling hard by calling a model with 30b parameters "nano"? With a bit of context, it doesn't even fit on a RTX 5090.
Other LLMs with the "nano" moniker are around 1b parameters or less.
FWIW It runs on my 9060xt(AMD) 16gb, without any tweaks just fine. It's very useable. I asked it to write a prime sieve in c#, started responding in .38 seconds, and wrote an implementation @ 20 tokens/sec
Getting ~150 tok/s on an empty context with a 24 GB 7900XTX via llama.cpp's Vukan backend.
Wow, Nvidia keepson pushing the frontier of misleading benchmarks
[dead]