Mistral Large 4 entered public preview on October 6, and its nickname, Le Chonk, is the least serious thing about it. The French lab describes a trillion-parameter model that reads images, targets coding, cybersecurity and finance, and will ship as open weights by the end of October. Some of that is documented. Some of it is Mistral’s own scorecard. Here is which is which.
What Mistral Large 4 actually is
Mistral’s docs list 1.05 trillion total parameters, with 49 billion active for each token. It is a mixture-of-experts design, so only a slice of the network runs per token. That keeps compute per request closer to a mid-sized model, but anyone hosting it still needs memory for every weight.
The model accepts images but answers in text only, according to Mistral co-founder Guillaume Lample, speaking to VentureBeat. Mistral says it trained the model from scratch on 3,800 Nvidia Grace Blackwell GPUs in its own European data centres. The preview API is open in Mistral Studio. Mistral calls it the first milestone funded by its €3 billion Series D from September.
The nickname comes from a June meme about a fictional giant Mistral model called Le Chaton Fat. Mistral’s CEO joked along. Then the company built a real one.
Confirmed vs not confirmed
✅ Confirmed
- Public preview launched October 6, 2026 (Mistral, TechCrunch, CNBC)
- 1.05T total and 49B active parameters (Mistral docs)
- Image input, text output only (VentureBeat)
- Weights promised by the end of October (Mistral)
⚠️ Not confirmed
- Weights on October 27 (VentureBeat only)
- License terms (one outlet expects a custom license, another says none announced)
- Context window: 1M tokens per Mistral, about 512K per one explainer
- Price: $1.36 / $4.18 in the launch post, $0.68 / $2.09 on the docs card
- Best open model from the US or Europe, plus coding and cyber leads: Mistral’s own claims, not yet on independent leaderboards
What Indian developers can do now
Start small and test with your own data. The preview API is open, but I could not confirm India availability or rupee billing, so check signup and billing before you plan around it. For budgeting, assume the higher price until Mistral clarifies, and see how other APIs compare in our AI API pricing comparison in rupees.
Test Indian languages yourself. Mistral says it trained on more than 160 languages but names only the EU ones. CEO Arthur Mensch told Akashvani in June that Mistral wants to improve Indian-language support, yet the launch materials contain no Indic benchmarks. Run real prompts in Hindi, Tamil, Bengali or whatever your users write, including mixed Hindi-English text. Compare accuracy and token counts against the model you use now, because Indic scripts can cost more tokens, and that changes your bill.
Think twice before self-hosting. By my own estimate, 1.05 trillion parameters is about 1 TB of weights at 8-bit, or roughly half that at 4-bit, before memory for context. That is a multi-GPU cluster, not a workstation. Our Kimi K3 self-hosting guide shows what that involves for another large open-weight model. Most startups will use an API or a hosted provider, and nobody should commit a product to the weights until the license is published.
Before you build on Mistral Large 4
What to watch next
Four things will settle most of the open questions: the weights release and its license, a stable price, a listing on independent leaderboards, and early Indic-language tests from developers. Until then, treat Mistral’s charts as promising, not proven.
Mistral Large 4: quick answers
Is Mistral Large 4 open source yet?
Not yet. Right now it is available only through Mistral’s preview API. Mistral promises open weights by the end of October, and VentureBeat reports October 27. The license has not been confirmed, and it decides how freely you can use the model commercially.
How much does the Mistral Large 4 API cost?
Mistral’s launch post lists $1.36 per million input tokens and $4.18 per million output tokens. Its docs card shows $0.68 and $2.09, and the reason for the difference has not been explained. Budget with the higher figures until it is.
Why is Mistral Large 4 called Le Chonk?
The name nods to “Le Chaton Fat”, a fictional giant Mistral model that went viral as a meme in June. Mistral adopted the joke as the nickname for the real model.
Sources: Mistral announcement (Oct 6, 2026), Mistral model docs (Oct 6, 2026), TechCrunch (Oct 6, 2026), VentureBeat (Oct 6, 2026), CNBC (Oct 6, 2026), Akashvani (Jun 18, 2026). Last Updated: October 7, 2026.