Nemotron 3 Ultra

NVIDIA · nvidia/nemotron-3-ultra-550b-a55b

← Back to leaderboard

NVIDIA Nemotron 3 Ultra is an open frontier-reasoning and orchestration model from NVIDIA, with 55B active parameters out of 550B total (MoE). Built on a hybrid Transformer-Mamba mixture-of-experts architecture, it...

Nemotron 3 Ultra ranks #248 of 348 models in the default AI Benchmark Hub order (quality first, then cost). At $1.88 blended per million tokens, it is cheaper than 36% of models with published API pricing. Its 256,000-token context window is about the catalog median (262,144 tokens).

open weightstexttext->text

Context

Max context: 256000
Max output: 32768

Pricing

Input / 1M: 0.63
Output / 1M: 3.13
Blend / 1M: 1.88

Quality

Quality index:

Provider

Provider: NVIDIA
Moderated: no

Monthly cost at this blended rate

Estimated API spend if every token is billed at $1.88per million (the blend of this model's input and output prices). Real bills depend on the input/output mix.

Tokens / monthEst. cost (USD)
1,000,000$1.88
10,000,000$18.75
100,000,000$187.50

Related models

Same lab, similar price, or similar quality — useful next comparisons.

Closest API price