Paris-based Mistral AI has released Large 4, a new large language model that carries the unlikely nickname “Le Chonk.” The name is a nod to a June internet joke, and it's already making waves: Large 4 outperformed GPT-6 Astra on one finance-specific benchmark, though it still trails Anthropic's Claude on other evaluations.
What's behind the 'Le Chonk' nickname
“Le Chonk” isn't a codename Mistral cooked up in a lab. It references a viral internet moment from June — the kind of meme that spreads across social media in days. Mistral, a company known for its irreverent streak, seems to have leaned into the joke. The nickname has stuck, and it's how the model is being talked about in early circles. The company hasn't explained the choice beyond the meme connection, but it fits Mistral's pattern of mixing serious research with a lighter public face.
The finance test where Large 4 beat GPT-6 Astra
On a single finance benchmark, Large 4 came out ahead of GPT-6 Astra, the latest from OpenAI. That's a notable win for a European challenger going up against one of the biggest names in AI. The test in question focuses on financial reasoning — exactly the kind of task that matters for banks, trading firms, and fintech startups. Mistral hasn't published a full breakdown of the scores, but the headline result is clear: on this one metric, Large 4 is on top. It's a narrow victory, though. One benchmark doesn't make a model, and Mistral's own results show there's plenty of ground left to cover elsewhere.
Where Large 4 still trails Claude
Against Anthropic's Claude, the story flips. Large 4 trails on a range of other tests, according to the available results. Mistral hasn't hidden that — the company's comparisons show Claude leading in several categories. That's not unusual for a new release; different models excel at different things. But it does mean Large 4 isn't a clean sweep. For users weighing which model to pick, the choice will depend on the job. If it's finance, Large 4 has a case. If it's general reasoning or other tasks, Claude still has the edge. Mistral hasn't specified exactly which benchmarks it lost on, only that Large 4 doesn't lead across the board.
Why a Paris lab matters in the AI race
Mistral AI is based in Paris, and that's not a trivial detail. The company has positioned itself as a European alternative to the U.S. giants — OpenAI, Anthropic, Google. Large 4 is its latest attempt to prove that a smaller, Europe-based team can compete on performance, not just on price or privacy. The finance benchmark win gives Mistral a talking point: it can beat GPT-6 Astra where it counts for some customers. The losses to Claude show the gap that remains. Mistral hasn't announced pricing or a release date for Large 4 beyond the launch itself. The model is out, the nickname is sticking, and the comparisons are already being picked apart by developers who want to see the full numbers.
What developers are watching next
The immediate question is how Large 4 performs outside the lab. Benchmark wins are one thing; real-world use is another. Finance firms that tested the model will want to see if the edge holds up on live data. Mistral hasn't said whether it plans to publish more detailed results or open up the model for wider testing. For now, Large 4 — or “Le Chonk,” if you prefer — is a reminder that the AI race isn't just about who has the biggest model. It's about who can win the specific tasks that matter. Mistral has one. Claude has others. The next round of comparisons will tell whether that split holds.
