Paris Mistral AI launched Mistral Large 4 on Tuesday, Oct. 6, introducing a 1-trillion-parameter AI model that the company now "very officially" calls Le Chonk. The model is priced at $1.36 per million input tokens and $4.18 per million output tokens, and Mistral says it will release the model's weights by the end of October.

Parameters are the adjustable numbers a model tunes during training — more of them generally means more capacity to learn, and higher running costs. Systems of this scale are the kind behind chatbots like ChatGPT and Claude.

A mixture-of-experts design

Not all of those parameters work at once. Large 4 uses a "mixture of experts" design, meaning the model routes each question to a few specialist sub-networks, so only 49 billion parameters fire per answer. Keeping active parameters low helps contain those running costs. Mistral's previous flagship, Large 3, worked the same way, with 41 billion active parameters out of 675 billion total.

"[Mistral Large 4] is at the frontier of open models, and by far the strongest open-weight model from the US or Europe," Mistral AI Chief Scientist Guillaume Lample wrote on X.

On Finance Agent v2, the launch chart shows Large 4 at 54.7 — ahead of GPT-6 Astra at 53.5, but behind Claude Opus 5.5 at 58.6.

Pricing undercuts rivals

Mistral bills by the token, the chunks of text — roughly three-quarters of a word each — that AI companies charge for. At $1.36 per million tokens and $4.18 per million output tokens, Large 4 lands at about a third of Claude Opus 5.5's input price and a fifth of its output price. Opus 5.5 charges $4 and $20, respectively, while GPT-6 Astra charges $10 and $50 — making Large 4 roughly a seventh and a twelfth of Astra's rates. Because per-token charges scale with usage, those gaps compound quickly for developers running heavy workloads.

How "Le Chonk" got its name

The nickname has a backstory. In June, after Mistral renamed its Le Chat assistant to Vibe, fans on Reddit and X invented a fictional model called Le Chaton Fat — roughly "the fat kitten" — with 30 trillion parameters, 1,000 meows per second, and fake benchmarks claiming it beat Claude Fable 5.

The French Goverment needs to STOP Le Chaton Fat before it's too late. This is the FAT takeoff we've been warned against for years! pic.twitter.com/VoaBPJmfBT

— GLIF (@heyglif) June 15, 2026

CEO Arthur Mensch played along, replying that the model was really called "le gros chaton," French for "the big kitten." Mistral then added a cartoon cat to its Vibe website. Its real launch post now lists the model as, "very officially," Le Chonk.

Sovereign AI and fresh funding

Mistral sells "sovereign AI" — models a country or company can own and run without handing its data to outside firms. Saudi Arabia's state-backed HUMAIN signed a deal worth hundreds of millions of euros with Mistral in August for exactly that reason. The planned weight release serves the same pitch: published weights let customers run the model on infrastructure they control.

In September, Mistral raised a €3 billion ($3.37 billion) Series D — a funding round in which a company sells shares to investors — at a valuation above €21 billion ($23.6 billion), in a round led by Samsung. The company says Large 4 is the first milestone on the roadmap that money funds.

What the benchmarks say

Mistral's announcement mostly benchmarks Large 4 against Chinese open-weight models — DeepSeek V4 Pro, Kimi K3, GLM-5.3, and Qwen3.8 Max. "Open-weight" means anyone can download and run the model. Claude and GPT models appear in only a handful of comparisons, and the newest Claude, Opus 5.5, shows up only in a cybersecurity claim. Launch-post charts are vendor-compiled, which is why third-party scores tend to draw extra attention.

In a blind human evaluation of coding quality run by Surge AI, Mistral Large 4 ranked second among five models with 3.74 out of 5, behind Claude Opus 5's 4.22.

AutomationBench, run by Artificial Analysis, hands an AI 657 chores in simulated business software — finance, HR, sales, and support — and scores the share of each task's goals it completes, with zero credit if it breaks a rule. Large 4 scored 59.9 points. On the same benchmark, Claude Sonnet 5.5 hit 71.8, Opus 5.5 hit 69.5, and Gemini 4 Argon topped the board at 77.5.

On DeepSWE 1.1, a benchmark developers consult when assessing how well a model codes, Large 4 scored 62 — ahead of GLM-5.3 at 61 and DeepSeek V4 Pro at 57, but trailing Kimi K3's 68. Datacurve's own leaderboard places GPT-6 Astra and Claude Opus 5 at 74.

Mistral says it will release Large 4's weights — the trained numbers that make the model work — by the end of October, which would let outside developers download the model and test those claims themselves.