October 6, 2026
Mistral Large 4 beats GLM-5.3 on coding benchmark with 62% on DeepSWE
According to VentureBeat as of Oct 6, Mistral Large 4 scored 62% on DeepSWE and outperformed GLM-5.3 in the comparison presented.
🚨 AI News | TestingCatalog
@testingcatalog
Mistral Large 4 scored 62% on DeepSWE and beat GLM-5.3, according to VentureBeat. It also scored 67% on Finch: finance tasks, the best result among open-weight models. Le Chonk also scored 15% on Harvey’s Legal Agent Benchmark: legal tasks, the best result among open-weight models. We need a technical report, right now 👀 Thanks for the tip @AiBattle_
BREAKING 🔥: Mistral announced Mistral Large 4 "Le Chonk", a new 1T-parameter open-weight model! > 49B active parameters, native multimodality. > Rolling out via APIs today; open-weight release is planned for the end of October. > SOTA on "critical workloads", including cyber defense. Le Chaton Fat "Le Chonk" is here 👀 https://x.com/MistralAI/status/2107457414387622310/video/1
· 12.6K views
Mistral has begun rolling out Large 4 through its API. The company plans to release the open weights in late October.
Coding performance. DeepSWE evaluates how well models complete software engineering tasks. In the comparison shown, Large 4 outperformed GLM-5.3, so the result concerns agent tasks rather than generating individual code snippets.
Finance and law. The model scored 67% on Finch and 15% on Harvey’s Legal Agent Benchmark. Both scores are claimed to rank first among open-weight models.
Original source: [Mistral Large 4 announcement](https://x.com/MistralAI/status/2107457414387622310).
Mistral plans to release the weights for Large 4 in late October 2026.