Mistral’s 1T-parameter ‘Le Chonk’ model: what’s reported and what’s confirmed

Mistral’s 1T-parameter ‘Le Chonk’ model: what’s reported and what’s confirmed

A reported trillion-parameter open-weight release lines up with a community meme, while Mistral's confirmed lineup tells a smaller, more practical story

A report from VentureBeat says Mistral has announced a one-trillion-parameter open-weight model. It is called ML4, nicknamed “Le Chonk,” and is reportedly open for testing ahead of a public release on October 27.

That would be a big deal for Europe’s best-known AI lab. The catch is that Mistral AI has not confirmed the existence of any such model.

The report and the meme

According to VentureBeat, ML4 carries one trillion parameters and ships as open weights. That means anyone could download the model and run it on their own hardware.

Advertisement

The ML4 and “Le Chonk” details align with an internet meme called “Le Chaton Fat,” which was created in June 2026. That meme is built around deliberately outlandish specs for a fictional Mistral model. Members of Mistral’s community have spread it with obvious relish.

CEO Arthur Mensch has reportedly joined in on the humor too. The specifications have been confirmed to be a joke rather than a product plan.

Mensch did hint at an “exciting” open-weight model for summer 2026. As of now, Mistral has not shipped a general-purpose model at the 1T scale.

What Mistral has actually released

Mistral Small 4 has 119 billion total parameters, with approximately 6.5 billion active at any given time. Small 4 supports a 256k context window and is multimodal, combining reasoning and coding capabilities in one package. It ships under Apache 2.0 open weights.

Mistral has also released Leanstral 1.5, which focuses on formal verification. That is the discipline of mathematically proving that a piece of reasoning or code does what it claims.

The money behind the models

In September 2026, the company closed a €3 billion Series D funding round. The deal valued Mistral at more than €21 billion. Samsung led the round, with participation from a16z and Nvidia. The capital is earmarked for research and infrastructure.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.
Mistral’s 1T-parameter ‘Le Chonk’ model: what’s reported and what’s confirmed
Mistral’s 1T-parameter ‘Le Chonk’ model: what’s reported and what’s confirmed

A reported trillion-parameter open-weight release lines up with a community meme, while Mistral's confirmed lineup tells a smaller, more practical story

A report from VentureBeat says Mistral has announced a one-trillion-parameter open-weight model. It is called ML4, nicknamed “Le Chonk,” and is reportedly open for testing ahead of a public release on October 27.

That would be a big deal for Europe’s best-known AI lab. The catch is that Mistral AI has not confirmed the existence of any such model.

The report and the meme

According to VentureBeat, ML4 carries one trillion parameters and ships as open weights. That means anyone could download the model and run it on their own hardware.

Advertisement

The ML4 and “Le Chonk” details align with an internet meme called “Le Chaton Fat,” which was created in June 2026. That meme is built around deliberately outlandish specs for a fictional Mistral model. Members of Mistral’s community have spread it with obvious relish.

CEO Arthur Mensch has reportedly joined in on the humor too. The specifications have been confirmed to be a joke rather than a product plan.

Mensch did hint at an “exciting” open-weight model for summer 2026. As of now, Mistral has not shipped a general-purpose model at the 1T scale.

What Mistral has actually released

Mistral Small 4 has 119 billion total parameters, with approximately 6.5 billion active at any given time. Small 4 supports a 256k context window and is multimodal, combining reasoning and coding capabilities in one package. It ships under Apache 2.0 open weights.

Mistral has also released Leanstral 1.5, which focuses on formal verification. That is the discipline of mathematically proving that a piece of reasoning or code does what it claims.

The money behind the models

In September 2026, the company closed a €3 billion Series D funding round. The deal valued Mistral at more than €21 billion. Samsung led the round, with participation from a16z and Nvidia. The capital is earmarked for research and infrastructure.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.