# Nemotron 3.5 Bolt 24GB



NVIDIA



An open multimodal model from NVIDIA built for local agents. Its experts use native 2-bit quantization so the model fits in 24GB of memory and runs alongside other apps, aimed at tool calling, web browsing and terminal tasks.


## Availability

Open weights on Hugging Face \(24GB 2-bit, GGUF, BF16; OpenMDW-1.1 license\); optimized for RTX Spark and RTX 5090
- Model released: Oct 15, 2026 · Date only
- Review updated: Oct 16, 2026 16:12 Fri · Beijing Time (UTC+8)

## Sources

- Source excerpt: https://x.com/NVIDIAAI/status/2110824076218605922
- Source archive: https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-24GB

Canonical entry: https://www.ailyre.com/source/?id=nvidia-nemotron-3-5-bolt-24gb&lang=en