Nemotron 3 Ultra
Nemotron 3 Ultra
Nvidia's largest open model, released with its training data as well as its weights.
Updated
The read
Nemotron 3 Ultra is the top of Nvidia's open family: a hybrid Mamba and transformer sparse model of 550 billion parameters with 55 billion active and a million-token window. Nvidia published the base and post-trained checkpoints, quantised versions and the datasets behind them, which few labs at this scale do. It is built for long-running agents on Nvidia hardware, and on general reasoning it sits in the middle of the open field rather than at the top.
Where it shines
- Weights, checkpoints and training datasets all published
- Million-token context on a hybrid design built for throughput
- Licence permits commercial use
Worth knowing first
- Mid-pack on reasoning against the leading Chinese open models
- Needs a multi-GPU server to run at full size
Index score
78.0
Provisional. This score is read from the product's public capability, not from the one-prompt rebuild the fully reviewed entries went through, so treat it as a placing rather than a verdict.
- Rank
- 51 of 73
- Builds
- LLMs
- Category
- Open weights
- Output
- Open weights and training data, commercially usable
- Pricing
- Free tier
- Best for
- Long-running agents on Nvidia hardware you already own
Scores are our own editorial judgement, weighted craft 55, speed 10, control 15 and value 20.
Also worth a look
Full index- 05GLM-5.3 FlashGLM-5.3 FlashProvisionalA fast, MIT-licensed open model that human voters rank beside the big flagships.Open weightsFree tier89.6
- 06Xiaomi MiMo V2.6 ProMiMo V2.6 ProProvisionalThe strongest large open-weights model on the independent index, under MIT, for under a dollar out.Open weightsFree tier89.4
- 07DeepSeek V4.1 FlashDeepSeek V4.1 FlashProvisionalFrontier-adjacent answers for thirty cents per million input tokens, with the weights under MIT.Open weightsFree tier88.3
