Mistral Large 4, aka 'Le Chonk': The French Heavyweight Against GPT-6 Astra and Claude

Science without conscience is but the ruin of the soul. An AI model named after a meme cat, but armed with 1 trillion parameters. Mistral unveiled Large 4 on October 6, unofficially dubbed "ML4" and very officially "the Chonk" according to its own announcement. The Parisian startup, which raised €3 billion in September, claims to be at the cutting edge among open models and asserts a slight lead over GPT-6 Astra in finance, with Claude still ahead. The figures come from the publisher, so they should be taken with a grain of salt.

Key points of this article:

• Mistral has revealed its AI model ML4, nicknamed "the Chonk," with 1 trillion parameters and a promise of European digital sovereignty.
• According to the publisher's figures, ML4 slightly outperforms GPT-6 Astra on a financial test but lags behind Claude and other competitors in other areas.

Large 4: What Mistral Announces

According to Mistral, ML4 is a multimodal model with a "mixture of experts" architecture (a set of subnetworks where only a portion activates for each request): 1 trillion parameters in total, with 49 billion active. Trained from scratch on about 4,000 Nvidia Grace Blackwell GPUs in Mistral's European data centers, in about two months according to VentureBeat, it covers more than 160 languages, including all official languages of the European Union. A preview version is available via the API. The model weights are expected to follow by the end of October.

Sovereignty is the main selling point. Mistral speaks of a European deployment operated end-to-end, "independently of other digital service providers and under European law." For a bank or a fund subject to GDPR, this argument carries weight. VentureBeat specifies that the weights will be released on October 27 under a proprietary license, which Mistral did not detail in its announcement.

Meet Mistral Large 4, aka Le Chonk.

• 1T parameters, natively multimodal. 49B active.

It is the best open weights model from the US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing, and finance, and it surpasses... pic.twitter.com/Jr5Z11mxEv
--- Mistral AI (@MistralAI) October 6, 2026
"Here is Mistral Large 4, aka Le Chonk. 1 trillion parameters, natively multimodal, 49 billion active. It is the best open weights model from the US or Europe on aggregated benchmarks. It is state-of-the-art on critical uses, including cyber defense, industry, and finance [...]"

Mistral AI, on X

Benchmarks: Ahead of GPT-6 Astra in Finance, Behind Claude Elsewhere

In finance, Mistral claims that ML4 surpasses GPT-6 Astra on legal and financial tasks. According to Decrypt, the Finance Agent v2 test gives 54.7 to Large 4 against 53.5 for GPT-6 Astra. Claude Opus 5.5, however, is ahead at 58.6. A gap of 1.2 points on a test provided by the publisher does not decisively separate anyone.

Elsewhere, the competition remains ahead. On the human evaluation of code quality by Surge AI, ML4 scores 3.74 out of 5 compared to 4.22 for Claude Opus 5. On AutomationBench (657 business software tasks), it scores 59.9 against 71.8 for Claude Sonnet 5.5. In DeepSWE 1.1 coding, it scores around 62 compared to 68 for Kimi K3.

Pricing and Finance: The Wallet Argument

The listed price is $1.36 per million tokens in input and $4.18 in output. According to Decrypt, this represents about a third of the entry price of Claude Opus 5.5 and a fifth of its output price. For massive uses, such as report analysis or credit file processing, the bill matters as much as the score.

The publisher also mentions the possibility of a private cloud or on-premises deployment, a sensitive point for financial data. Results remain to be confirmed by independent tests after the release of the weights. The startup, valued at over €21 billion after its September round, is betting its credibility on this meeting.

Large 4 arrives in a landscape where European digital sovereignty is at stake as much in hardware as in models, and Mistral still relies on Nvidia chips for training. The first independent measures are expected after the publication of the weights on October 27, 2026.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.