Skip to main content
Mistral•medium Impact•

Mistral Large 4: Public Preview Now, Reduced Moderation for a Vetted Group

News by OneHuman

Mistral Large 4 launches as a public preview while vetted partners get reduced moderation. A third case of access split by audience; weights due by month end.

mistralmistral-large-4model-availabilityaccess-tiersconsumer-protectionoctober-2026

Published: October 7, 2026 Impact: Medium. Affects anyone judging a Mistral model by what the public tier does.

On October 6, 2026, Mistral launched a public preview of Mistral Large 4. A vetted group gets "the same model with reduced moderation and expanded cyber capabilities." That makes Mistral a third lab, after Anthropic and Google, to make access to a new model depend on who you are.

The public can use Large 4 today; the version with reduced moderation is reserved for the groups Mistral names.

What Happened (This Week)

Mistral's launch post, dated October 6: "Today, we're launching a public preview of Mistral Large 4." And: "You can try the preview API today on Mistral Studio. Weights drop end of this month."

Mistral says Large 4 has 1 trillion parameters, 49 billion active. The page's model card lists API pricing of $1.36 per million input tokens and $4.18 per million output tokens.

On access, Mistral writes: "We will release the weights by the end of the month. Until then, we are red-teaming the model in real-world settings with cybersecurity leaders, vetted partners, and state authorities, who will access the same model with reduced moderation and expanded cyber capabilities."

What This Means

OneHuman's October 3 piece called access-restricted releases "n=2, a co-occurrence, not a pattern" and listed whether they stay at n=2 as the watch item. Mistral is a third case, with a difference.

Mythos 5.1 and Argon were not generally usable by the public at launch. Large 4 is: the public can try the preview API on Mistral Studio today. Mistral also gives a deadline for its next step, weights by the end of October. Google's Argon post gives no date for wider release, and Anthropic's page says its Cyber Verification Program will include Mythos-class access "in the near future." So this is not a third restricted-first release. It is a third case of audience-dependent access, and the closest match is Anthropic's Mythos and Fable: the same model at different safeguard levels.

Counter-evidence stands. Fable 5.1 is the same model as Mythos 5.1 and is generally available. Three cases is still a small set.

Mistral's brand is open weights. OneHuman's read: the weights are the part to watch. The launch post does not say which safeguard level the weights will carry.

Consumer Protection Q&A

Q: Can I use Mistral Large 4 today? A: Yes, as a public preview through the API on Mistral Studio, at $1.36 input and $4.18 output per million tokens. The page does not mention the Vibe app, so subscribers should not assume it is there.

Q: Is the vetted tier a different model? A: Mistral says the vetted group accesses "the same model with reduced moderation and expanded cyber capabilities." Same model, different moderation.

Q: Will the open weights carry the public or the vetted safeguards? A: The launch post does not say. It says Large 4 "pairs top-tier cyber performance with open weights and self-deployment, giving organizations both the capability and the autonomy to run advanced security work under their own policies." Whoever self-deploys sets their own moderation.

Q: When do the weights arrive? A: Mistral says "by the end of the month," with no exact date.

Q: How does the price compare with Mistral's other models? A: Large 4 lists below Mistral Medium 3.5's $1.50 input and $7.50 output on the Compare Page.

What Happens Next

  • 30 days: Do the weights arrive by October 31, and with which safeguards?
  • 90 days: Does a fourth lab make access depend on the audience?
  • 6-12 months: Do the self-deployed weights become the version people test, rather than Mistral's hosted preview?

What to Do Next

  • Check which tier you are on before judging the model. The vetted tier is limited to the groups Mistral names.
  • Vibe subscribers: the launch page does not mention the app, so check before assuming access.
  • If you plan to self-host, wait for the weights and read their terms.

Bottom Line

What's real: one model, two safeguard levels, split by audience. What to watch: whether the weights match the public tier or the vetted one. What's next: the access tier is part of the product.

Sources

Verified by OneHuman · October 7, 2026

Independent. AI-assisted. Human-verified. No ads. No affiliates. No investors.

Share This Article

"Mistral Large 4 is a third case of access depending on the audience. The public gets the preview. A vetted group gets the same model with reduced moderation."
— News by OneHuman
"Mistral gives a deadline for its next step: weights by end of October. Google's Argon post gives no date, and Anthropic says Mythos-class access comes only 'in the near future.'"
— News by OneHuman
"Mistral's brand is open weights, yet its newest model launches with a public preview and a vetted tier with reduced moderation. Check which tier you are on before judging the model."
— News by OneHuman
"Three cases is still a small set. The access tier and safeguard level are part of the product, so read them the way you read the price, before you judge what a model can do."
— News by OneHuman

Join OneHuman to keep Mistral, and every AI tool you pay for, honest.

Independent. AI-assisted. Human-verified. No ads. No affiliates. No investors.

No AI lab, including the one behind Mistral, pays for or influences this coverage. Why this matters →

Join OneHuman (Free or Pro) →

Pro unlocks unlimited recommendations and Pro Deep Analysis.