Mistral AI has released Mistral Large 4 (ML4), a trillion-parameter Mixture of Experts model that aims to compete with leading AI models from the US and China. Independent tests by Artificial Analysis show ML4 performing strongly in cybersecurity, outperforming several Chinese models in specific tests, though it trails some Chinese open-weight models on broader intelligence benchmarks. The model, trained on Nvidia Grace Blackwell GPUs, offers a 1 million token context window and native image input, with its weights expected to be released by the end of October. AI
IMPACT Positions Mistral AI as a strong contender against US and Chinese AI labs, particularly in specialized domains like cybersecurity.
RANK_REASON Frontier-lab model release with system card.
Read on Mastodon — fosstodon.org →
- Le Chonk
- Mistral AI
- Mistral Large 4
- Nvidia
- Anthropic
- Artificial Analysis Cyber Index
- China
- Claude Opus-5
- Cybench
- CyberGym-E2E
- GLM 5.3
- Kimi K3
- Nvidia Grace Blackwell
- OpenAI
- Arthur Mensch
- Artificial Analysis
- Claude Opus 5.5
- CyberGym-E2E-AA
- DeepSeek V4.1 Flash
- GPT-6 Luna
- MiMo-V2.6-Pro
AI-generated summary · Google Gemini · from 12 sources. How we write summaries →