NVIDIA has open-sourced its Nemotron 3.5 Bolt 74B-A9B, a 74-billion-parameter agent model designed to run on consumer hardware with 24GB of memory. Unlike recent closed-source releases from Meta, OpenAI, and Google, NVIDIA is providing the model weights, training data, and recipes for free download. This open approach allows users to fine-tune the model for offline use and deeper inspection, though independent benchmarks are still pending. AI
IMPACT Enables offline AI agent development and inspection, challenging the API-based model of closed-source competitors.
RANK_REASON Open-source release of a large (74B parameter) agent model by a frontier lab (Nvidia) with detailed technical specifications. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
- Gemini 4 Argon
- GPT-6
- Hugging Face
- Meta*
- Muse Wonder
- Nemotron 3.5 Bolt 74B-A9B
- Nvidia
- Ollama
- OpenAI
- RTX Spark
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →