NVIDIA / nvidia/nemotron-3-ultra-550b-a55b-262k

Nemotron 3 Ultra 550B A55B 262K - access through LLMTR

Nemotron 3 Ultra 550B A55B is NVIDIA's largest open-weight Nemotron model, a sparse Mixture-of-Experts architecture that runs 55 billion of its 550 billion parameters per token. It is built for frontier reasoning, orchestration, coding agents and long-running enterprise workflows. This deployment offers a 262,144-token context window, native tool calling, JSON object mode and JSON Schema structured output. Use `tool_choice` with `auto` or `required`; forcing one named function is not enforced here — the model answers with whichever tool fits the prompt instead of the one you named, so do not rely on a named choice to constrain it. Prompt caching is applied automatically: a repeated prefix is retained in whole blocks of 8,192 tokens and billed at a discounted cached-input rate, while the remainder bills as standard input. Step-by-step reasoning is on by default and returned separately from the answer, so even a short question spends reasoning tokens and you should size `max_tokens` accordingly; it can be switched off per request, and reasoning tokens bill as output. The model accepts text only and produces text only. There is no separate output ceiling below the context window.