Nemotron 3 Ultra

Nvidia · Nemotron 3 · open · released Jun 4, 2026 · nemotron-3-ultra-550b-a55b

Open frontier-reasoning and orchestration model from Nvidia: a hybrid Transformer-Mamba mixture-of-experts with 55B active of 550B total parameters and a 1M-token context window.

Pricing

Input$0.60 / 1M checked Jul 19, 2026
Output$3.60 / 1M checked Jul 19, 2026

Hosted-inference pricing as listed; open weights available for self-hosting.

Cost bandBudget
Free tierOpen weights published; hosted API pricing shown.

Limits

Context window1M checked Jul 19, 2026
DeploymentSelf-hosted, NVIDIA NIM, partner inference
OpenAI-compatible APIYes
ModalitiesText

Where it fits

Best forSelf-hosted frontier reasoning and agent orchestration on Nvidia hardware.
Business fitOpen weights with hosted pricing under $4 per 1M output tokens.

Capabilities

ReasoningFunction calling

Change history

Data

Last checkedJul 19, 2026
Cross-checked againstopenrouter.ai/nvidia/nemotron-3-ultra-550b-a55b

Verify with provider documentation before committing spend. Report an error and it gets fixed with a logged correction.

Open in the comparison table JSON API