N
Nemotron 3 UltraOPEN
NVIDIA Nemotron 3 Ultra is an open frontier-reasoning and orchestration model from NVIDIA, with 55B active parameters out of 550B total (MoE).
Prices update automatically — checked daily against provider list prices.
- $1.05 per 1M tokens, assuming 3 input for 1 output the convention used by Artificial Analysis
- cheaper than 14% of the 200 open-weight models tracked here
- 12 of 12 NVIDIA models tracked here, from cheapest
Benchmarks & performance
Independent benchmark scores for Nemotron 3 Ultra, measured by Artificial Analysis. Higher is better. Measurement mode: Reasoning.
Models at this level cost $0.64 per 1M tokens (median of 5) — this one costs $1.05.
Cheaper and better on this index: GPT-6.1 Terra (batch) · Claude Haiku 5.5 · Claude Haiku 5.5 (batch) and 15 more
About this model
Nemotron 3 Ultra is an open-weight AI model by NVIDIA. You can download and self-host it for free; the prices below are hosted-API list prices, tracked daily, for when you prefer convenience over self-hosting.
Frequently asked questions
What is Nemotron 3 Ultra?
Nemotron 3 Ultra is an AI language model from NVIDIA. It is open-weight: you can download it and run it on your own hardware, for free. It scores 11.8 on the Artificial Analysis intelligence index.
Is Nemotron 3 Ultra free?
The weights are free and open — you can self-host Nemotron 3 Ultra and pay nothing per token. If you prefer a hosted API, list prices are $0.6 per million input tokens and $2.4 per million output tokens.
What is Nemotron 3 Ultra good at?
Independent benchmarks from Artificial Analysis give it GPQA 86.7%, Humanity's Last Exam 28.4%, Long Context Reasoning 79.3%, SciCode 40.3%, IFBench 81.4%, τ²-Bench 83.3%, τ-Bench Banking 14.2%, Terminal-Bench 53.9%, Terminal-Bench Hard 36.4%. It is particularly used for code generation.
How fast is Nemotron 3 Ultra?
It generates about 173.4 tokens per second, with a median 0.9s delay before the first token. Measured independently by Artificial Analysis.
Can I self-host Nemotron 3 Ultra?
Yes. Nemotron 3 Ultra has open weights, so you can download it and run it on your own GPU or server with tools like Ollama, vLLM or llama.cpp — with no per-token cost.