Nemotron Nano 3 30B-A3B
20 GB • 1.0m ctx • Q4_K_M
← All models4K 25 GB 8K 26 GB 16K 28 GB 33K 30 GB 66K 31 GB 131K 33 GB 1.0M 38 GB
NVIDIA2025
Nemotron Nano 3
Hybrid MoE · 1M context
NVIDIA hybrid MoE for agents: reasoning, coding, and tools with 1M-token context and high throughput. The 30B-A3B active path stays efficient while the long window holds entire projects.
- Fits fromWorkstation · 32 GB+
- Contextup to 1.0M
- Sizes1
- RunsLocal on Delta
AgentsReasoningCodingLong-horizon agentsCoding1M context
Sizes
Memory at context length
Approximate working set for Nemotron Nano 3 30B-A3B as context grows.
