Automate Basics
New model · TechCrunch

Mistral releases Large 4 with access limited during safety testing

By Automate Basics, written with AI from TechCrunch's original

· 2 min read

Hand sorts printed financial charts on a meeting-room table in daylight
Image: AI-generated illustration.

Mistral AI has released Large 4, a multimodal model that is currently accessible through a guarded endpoint. The company plans to make its model weights available after safety testing, a distinction for teams weighing how much control they need.

What changed with Mistral Large 4?

Mistral AI has released Mistral Large 4, a multimodal AI model. TechCrunch reports that people can currently access it only through a public endpoint with safeguards. Mistral has not yet made the model weights available for others to run themselves.

Mistral plans to release those weights in three weeks, after safety testing is complete, according to TechCrunch. That plan matters to organizations that want to examine or run a model more directly, rather than rely solely on a provider's hosted service. For now, that choice is not available. Mistral identifies cybersecurity, finance and chip design as areas it has focused on, but benchmark results were still pending when TechCrunch reported the release.

Who can use Mistral Large 4, and what does it cost?

Mistral Large 4 is currently available through a public guarded endpoint, TechCrunch reports. The report does not say which plans include access, what using the endpoint costs or whether every organization can use it under the same terms. Teams considering it for work should check those details before budgeting for a trial.

The model may be of particular interest to organizations assessing AI for specialized work. Mistral points to cybersecurity, finance and chip design as intended areas of strength. Those are the company's priorities, not independently established performance results. Teams that need access to model weights should also note the difference between a planned release and one they can use today. TechCrunch says safety testing must finish before Mistral makes the weights available.

How do I try Mistral Large 4 at work?

A useful first trial for Mistral Large 4 would be a small, low-risk task from an area Mistral has highlighted, such as explaining a financial chart or summarizing a non-sensitive security report. TechCrunch does not provide instructions for accessing the guarded endpoint or specify which kinds of files it accepts, so confirm the available input options before choosing a task.

Compare the response with material your team already trusts. Check whether it identifies the main facts, handles uncertainty clearly and avoids adding claims that are not in the source material. Do not treat a plausible answer as proof that the model performs well in a specialist field. TechCrunch says benchmark results were still pending, and its report does not establish how the model performs on a particular workplace task.

Frequently asked questions

Are the Mistral Large 4 model weights available?

No. TechCrunch reports that Mistral Large 4 is currently accessible only through a public guarded endpoint. Mistral plans to make the model weights available in three weeks, once safety testing is complete. That is a stated plan, not access that teams have today.

Is Mistral Large 4 proven to be better for finance or security work?

Not on the evidence in TechCrunch's report. Mistral identifies finance and cybersecurity among the areas it has focused on, but benchmark results were still pending. A team considering the model should check its answers against trusted work material before relying on them.

Written with AI from TechCrunch's original and published after automatic checks: every figure here appears in the original, and no sentence is copied from it. The picture is AI-generated. The original is the authority.

Source: Mistral’s new 1T model aims to leapfrog closed and open rivals, TechCrunch, 6 Oct 2026.