NVIDIA: Nemotron 3 Ultra (free)

Nemotron 3 Ultra is a text-only model from NVIDIA offering a 1-million-token context window and up to 65,536 tokens of completion output. It supports tool use and reasoning, which makes it relevant for agentic workflows and multi-step tasks. Structured output support is unconfirmed, so builders who depend on reliable JSON formatting should verify that before committing. The model is free, which removes cost as a barrier for experimentation or high-volume use cases. The tradeoff is that it currently has no independent benchmark coverage, so there is no public data to gauge where it stands relative to paid alternatives. It is a reasonable shortlist candidate for developers exploring reasoning-capable models on a zero budget, but anyone making a production decision should treat its quality as unproven until external evaluations become available.