Announcing native availability of NVIDIA Nemotron 3 Nano, NVIDIA’s latest reasoning model
Together AI
NVIDIA released Nemotron 3 Nano, a reasoning model with a hybrid Mamba-Transformer architecture that activates approximately 3 billion parameters per token out of 30 billion total, now available on Together AI. The model includes 1 million token context length, open weights, and open training data optimized for agentic systems and multi-agent applications. The availability enables developers to build cost-efficient reasoning agents and multi-step planning systems with faster inference speeds compared to larger models.
Why it matters
Nemotron 3 Nano, NVIDIA’s newest reasoning model, is now available on Together AI, the AI Native Cloud
