Llama-3.1-Nemotron-Ultra-253B
Nemotron Ultra, NVIDIA's flagship Nemotron tuned over Llama 3.1. 253B parameter dense model targeting frontier-class quality.
Llama-3.1-Nemotron-Ultra-253B is a frontier AI model from Nvidia. It costs $1.60 per million input tokens and $1.60 per million output tokens (blended $1.60/M), with a 128,000-token context window.
RatesUSD per million tokens. Blend is 70% input, 30% output, our calculation.
Hand-maintained from Nvidia's published pricing; no check date on record.
- Largest Nemotron
- Open weights
- NVIDIA NIM hosting
- Frontier-class self-hosted inference
- Synthetic data generation
Benchmarks
Also reported by the provider
Results on benchmarks outside the set we compare across models. They are listed for completeness and are deliberately excluded from the bars above and from every ranking on this site, because a score is only comparable against the same benchmark variant. Why benchmark scores do not compare →
- MATH-500 (pass@1, reasoning on)97%self-reported by NVIDIA · 2025-03 · Reasoning on (reasoning off: 80.40). 32k sequence length, averaged over up to 16 runs, temperature 0.6, top_p 0.95. Not comparable with the MATH column, so shown here and never ranked.
More from Nvidia
See all 6 →Frequently asked questions
How much does Llama-3.1-Nemotron-Ultra-253B cost?
Llama-3.1-Nemotron-Ultra-253B costs $1.60 per million input tokens and $1.60 per million output tokens, for a blended reference rate of $1.60 per million tokens.
What is Llama-3.1-Nemotron-Ultra-253B's context window?
Llama-3.1-Nemotron-Ultra-253B supports up to 128,000 tokens of context in a single request.
What is Llama-3.1-Nemotron-Ultra-253B best for?
Llama-3.1-Nemotron-Ultra-253B is well suited to Largest Nemotron, Open weights and NVIDIA NIM hosting.
Who makes Llama-3.1-Nemotron-Ultra-253B?
Llama-3.1-Nemotron-Ultra-253B is developed and served by Nvidia.
