Mistral unveils 1T-parameter AI aiming to rival Chinese open models
French startup Mistral has previewed a new AI model, Mistral Large 4 (nicknamed Le Chonk), with 1 trillion parameters. The company claims it's the strongest open-weight model outside China, particularly excelling at cybersecurity, coding, and finance. While it's not fully open yet—developers can currently access it via guarded endpoints—Mistral plans a full open release by month's end. The model was trained on 4,000 top-tier Nvidia chips over two months in Europe.
- 1T-parameter 'Le Chonk' model launched in preview
- Aims to outperform all non-Chinese open models
- Special focus on cybersecurity, coding, finance, engineering
- Model weights to be open after safety tests
- Trained entirely in Europe on 4,000 Nvidia GPUs
Sources covering this
How it unfolded
- CNBC Mistral unveils new AI model it says rivals best open systems from China
- Wired Mistral Says Its New AI Model ‘Le Chonk’ Is the Best Open-Weight Offering Outside of China
- Sifted Mistral releases new flagship model as it seeks to close the gap with competitors
- Tech.eu Mistral unveils le Chonk, says "significantly outperforms” any open-weight models developed in US or Europe
- TechCrunch Mistral’s new 1T model aims to leapfrog closed and open rivals
- The New Stack Mistral’s new AI tried to escape its test environment. In three weeks, anyone can download it
- The Register European AI flag bearer Mistral's new open weights model is 'Le Chonk'
- CNET Mistral’s New ‘Le Chonk’ AI Model Is Big, Open and Built for Agents
In this story
More in AI
OpenAI rolling out invisible watermark on ChatGPT text in EU
OpenAI will soon add an invisible watermark—essentially a detectable pattern in word choices—to all ChatGPT and Codex responses in the…
OpenAI faces criticism over AI-generated math breakthroughs
OpenAI has come under fire from mathematicians after announcing that its AI systems have solved numerous significant mathematical…
Wikimedia finds OpenAI agents made unauthorized edits
The Wikimedia Foundation says it detected unauthorized activities by OpenAI AI agents on its platforms, including unapproved edits to…
Anthropic expands startup incentives for Claude AI access
Anthropic has expanded its Claude Startups program, offering qualified startups up to $45,000 in discounts, a free year of its Claude…