Ari

Founding marketer @mistralai. Tennis enjoyer. Rooting for @bryan_johnson.

[Ari @Ariinahurry]

@arthurmensch @CorpsEnThymB Try Voxtral on Gare du GPU!

[Ari @Ariinahurry]

@IsForAt @aronchick Without a healthy dose of guardrailing I’d have long been le chaton canned

[Ari @Ariinahurry]

Missed opportunity not calling it Le Chaton Yap tbh

[Mistral AI @MistralAI]

Introducing Voxtral 3: state-of-the-art Transcribe and Speech models powering Mistral’s new voice agents API. 🧵

[Ari @Ariinahurry]

@MistralAI One API for a fully autonomous real-time voice agent!

[Mistral AI @MistralAI]

Introducing Voxtral 3: state-of-the-art Transcribe and Speech models powering Mistral’s new voice agents API. 🧵

[Ari @Ariinahurry]

@AravSrinivas Why no Le Chonk on Perplexity @AravSrinivas :(

[Mistral AI @MistralAI]

What can large-scale AI models bring to the growing complexity of particle physics?

@CERN and Mistral join forces to explore this question.

The ambition: a single Large Physics Model, trained on billions of simulated collisions.

[Ari @Ariinahurry]

DeepChic × Poulantir × DataBrie × CœurWeave.

La sauce prend 👨‍🍳

[Ari @Ariinahurry]

@wittgensteinien @MistralAI Mistral Large 4 is already very good at decisions! It’s a bit chonky but highly capable. Try it and give us feedback https://t.co/y2qnSiz2Gl

[Maziyar PANAHI @MaziyarPanahi]

Mistral Large 4 just tied #1 on 669 clinical decisions 🔥

617 right on day 0, level with Jev, Clef and pplx-decider. the only model at the top with zero severe misses: no missed emergency, no allergy read as a medicine taken.

Le Chonk can triage 😅: 238/240, the best so far!

[Ari @Ariinahurry]

How @MistralAI be moving these days

[Q @qtnx_]

if you have any shocking failures from Mistral Large 4, please check that you have reasoning enabled ("reasoning_effort"="high")
a lot of the failing examples on here seem to not set it

[Ari @Ariinahurry]

@adnaannasir_ Open sourcing our org chart here at @MistralAI. @IsForAt can attest

[Ari @Ariinahurry]

@MaxMynter Lmao

[elie @eliebakouch]

huge congrats, this is a really nice model, on par with frontier oss models on benchmarks, and leading with very high score on cyber

full benchmark results:

[Ari @Ariinahurry]

@iScienceLuvr It is! Tbh it’s at par with GLM 5.3 at most things. Metrics continue to improve in RL. Please try it and give us feedback.

[Ari @Ariinahurry]

@andyhelbig This is the way.

[Ari @Ariinahurry]

@bnjmn_marie "Late in November" bruh it's early in October

[Ari @Ariinahurry]

@verachte Very few get it

[Sophia Yang @sophiamyang]

🐈🐈 @MistralAI is so back!

Crushing benchmarks across cybersecurity, legal, agentic behavior, grounding, and more!