Mistral’s 1T-parameter Large 4 takes on both closed and open rivals

French AI lab Mistral AI has released Mistral Large 4 (ML4), a new large multimodal model with which it hopes to leapfrog both American and Chinese rivals. The company frames it as an embodiment of what French president Emmanuel Macron has called “a third way in AI.”

Amid a growing divide between closed models that can be unplugged and open models that are often made in China, Mistral is positioning ML4 as an alternative to both. Nicknamed Le Chonk for its one trillion parameters, the model is anything but small — yet it is not an open-weight model, at least not yet.

Access and open weights

For now, ML4 can only be reached through a public guardrail endpoint. Mistral plans to release the weights in roughly three weeks, once safety testing is complete. In the meantime, VP of Science Pierre Stock told TechCrunch, the company will work with trusted partners and governments to ensure the open weights can be used to defend, but not to carry out malicious attacks.

Security concerns have mounted in recent months, especially among Mistral’s core audience of enterprises and institutions. At the same time, Stock noted, an open-weight model is easier to audit.

Lean compute

Another point the company stresses is that ML4 was trained entirely on Mistral’s own compute, using just 4,000 NVIDIA GPUs — “which is two to three times less than our Chinese competitors, and significantly less than the closed source competitors,” Stock said.

Benchmark results are still pending, but Mistral hopes ML4 will be best in class among open-weight models, especially outside China, though not only there. Thanks to focused training, Stock argues, it could also outperform closed models in specific areas that matter to its customers.

Industry focus

Among its optimized use cases, Stock cites cybersecurity and finance, as well as chip design — a theme close to two of Mistral’s main backers. Those are Dutch giant ASML, which led its Series C, and Samsung, which led its Series D last month at a €21 billion valuation (about $24.39 billion).

Earlier, in explaining its decision to host Chinese models, Mistral had tried to signal that it was not becoming a mere inference provider. With Le Chonk in its corner, the company intends to keep being seen as a frontier lab.

Source: TechCrunch