Mistral’s Le Chonk is now a top open model for coding, ranking alongside Codex GPT-6.1 Terra
Mistral has published the weights for Mistral Large 4 (ML4), the trillion-parameter model it previewed on Oct. 6 and nicknamed "Le Chonk."
New benchmark scores show ML4's result on the Terminal-Bench 4.0 agentic coding benchmark rising from 28% in preview to 41%, one point ahead of GLM-5.3. On a composite index across three benchmarks, ML4 scores 48, one point behind GLM-5.3's 49 and level with Codex GPT-6.1 Terra.
ZDNet cautions that benchmark numbers should be taken with a grain of salt, but says the gain after only a few more weeks of training is meaningful and shows Western open models catching up with Chinese ones while proprietary AI stays expensive.
Key points
- Open weights released for ML4
- Terminal-Bench 4.0: 28% to 41%
- Index score 48, tied with Codex GPT-6.1 Terra
Summary written by Readthrough from the original article. The full story, with photos and any updates, is at ZDNet.
