Models · NVIDIA

Nemotron 3 Ultra (free)

Released 4 Jun 2026Tool useReasoningFree

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...

InputFreeper 1M tokens
OutputFreeper 1M tokens
Context1Mtokens it can read at once
Max output66Ktokens per reply

What members say

From members who used it for real work. Facts above come from OpenRouter.

No reviews yet. Used it? Tell others what it was good or bad at.

Members can review models they've used.