Nemotron-3-Nano 4B
2.8 GB • 1.0m ctx • Q4_K_M
← All models4K 3.5 GB 8K 3.7 GB 16K 4 GB 33K 4.2 GB 66K 4.4 GB 131K 4.7 GB 1.0M 5.3 GB
NVIDIA2026
Nemotron 3 Nano
Efficient dense · agentic
A compact NVIDIA agentic model built for local assistants. The 4B variant keeps reasoning and tool-use quality high while staying light enough for laptops. 1M-token context means long documents, logs, and codebases stay in one session, all on your hardware.
- Fits fromLaptop · 7 GB+
- Contextup to 1.0M
- Sizes1
- RunsLocal on Delta
AgentsReasoningEdgeLocal agentsLong-context chatLaptops
Sizes
Memory at context length
Approximate working set for Nemotron-3-Nano 4B as context grows.
