Mistral, the French AI company, has opened a public preview of Mistral Large 4, nicknamed “le Chonk” after an internet meme about a giant French cat model. It is the largest model Mistral has built: about 1 trillion parameters in total, of which only 52 billion run for any single token. It reads images yet answers in text, holds a 1 million-token context window, and was trained from scratch on 3,800 Nvidia Grace Blackwell chips in Mistral’s own European data centers, across more than 160 languages. The preview runs on Mistral’s API at $1.36 per million input tokens and $4.18 per million output. The full weights, which let anyone run the model on their own machines, are promised by the end of October.
Enterprises and governments want control: with downloadable weights, a security team can run ML4 in-house and scan code without a provider’s filter intervening. Mistral points to an Artificial Analysis test that asks a model to reproduce a real software flaw and then patch it. ML4 scores 82%, the highest of any model, while Claude Opus 5.5 and GPT-6 Astra score near zero because they refuse the task. A refusal there is exactly what stops a defender from proving a vulnerability is real.
Artificial Analysis’s independent index puts the preview at 38, mid-pack across 219 models, and the live DeepSWE coding leaderboard has open rivals GLM-5.3 and Kimi K3 near 69% and frontier models near 74%, above ML4’s 61.7%. The firmest test arrives when the weights land and outsiders can run it themselves.
Read More: Qwen3.6 and DeepSeek V4: China’s Open-Weight Models Now Match Frontier Competitors
Sources:
- Introducing Mistral Large 4 (Mistral)
- Mistral Large 4 model documentation (Mistral)
- Mistral Large 4 Preview: intelligence, performance and price analysis (Artificial Analysis)
- Mistral debuts Large 4 ‘Le Chonk’, a 1-trillion parameter text output model (VentureBeat)
- Mistral AI Releases Mistral Large 4 (Le Chonk): a 1.05T parameter multimodal MoE (MarkTechPost)
Disclaimer: For information only. Accuracy or completeness not guaranteed. Illegal use prohibited. Not professional advice or solicitation. Read more: /terms-of-service
Reuse
Citation
@misc{kabui2026,
author = {{Kabui, Charles}},
title = {Mistral {Large} 4 “Le {Chonk}”: A {Trillion-Parameter}
{Open-Weight} {Model} {From} {Europe}},
date = {2026-10-13},
url = {https://toknow.ai/posts/mistral-large-4-le-chonk-open-weight-trillion/},
langid = {en-GB}
}