mistralai/Leanstral-1.5-119B-A6B huggingface.co/mistralai/Lea…

Mistral's funding and "frontier AI for everyone" mission raises a real question: what does "democratized" actually mean when you need enterprise infrastructure to run it safely? Open source models are great. But most teams deploying them still need visibility into what data flows through their calls, cost controls, and compliance logging. The models democratized, the governance stack didn't. @MistralAI #AIGovernance #OpenSource #GenAI

Thank you. Excited to be here and looking forward to contributing.

Welcome to the program! Glad to have you!

#MistralAI lance #Leanstral1.5~? l'#IA gratuite qui vérifie l'exactitude de votre code~? un modèle d'agent de code #Lean4 #opensource capable de résoudre 587 des 672 problèmes du #PutnamBench dlvr.it/TTNB4C

🫶

AI News Roundup - Past 24 Hours Aggregated insights from the tech community MistralAI European AI Expansion Summary: MistralAI secured new funding for scaling European AI infrastructure and models. Could this balance global AI power dynamics? Source: x.com/MistralAI #AINews #5MinAINews

Congrats to @MistralAI on Leanstral 1.5! 🎉 An Apache-2.0 Lean 4 proof agent that punches way above its size: 🧩 MoE: 119B total, just 6B active 📐 100% on miniF2F 🎓 New SOTA on FATE-H (87%) & FATE-X (34%) ⚡ 587/672 on PutnamBench at ~$4/problem Read the blog below or serve it on vLLM today! huggingface.co/mistralai/Lea…

but now "european company", obviously I'm saying MistralAI and maybe Blackforest Lab is commiting suicide by strict censorship, yet bad performance. I won't expect russia will surpass China or US, since AI is still compute-hungry and power-hungry, but at least nobody will ignore.

I think @MistralAI vibe is pretty good, BUT you must never lot it come close to the context window size or do compaction while it is in mid-PR generation. That will go very, very wrong.

Stop calling Mistral AI "the French OpenAI." ❌ In 2026, they are winning by running the Palantir playbook: embedding engineers directly with enterprise and government clients to build custom models via their Forge platform. kalinga.ai/mistral-ai-guide-… #AI #MistralAI #TechNews

Counting down to @RaiseSummit where our team will show exactly what we’ve built: Brick. Brick reads every prompt's capability (6 dimensions: coding, math, reasoning, creative, planning, world knowledge) and complexity, then routes it to the cheapest model in your pool that can actually do the job. Benchmark results (n=5,504 queries): - Always Kimi2.6 → 75.02% accuracy at 6× cost - Brick max-quality → 76.98% accuracy at 1.5× cost - Brick max-saving → 72.4% accuracy at 1.0× cost Brick beats the strongest single model in both quality and cost. Works with: → Claude Code (brick claude on — done) → OpenAI Codex (brick codex on) → Any OpenAI-compatible client (model: "brick") → Your own model pool (config.yaml, any backend) No GPU needed to run the router. CPU box is enough if you use a hosted pool. GitHub: github.com/regolo-ai/brick-S… In Paris next week? Let’s meet and talk 👋 @RunPod @AnthropicAI @OpenAI @huggingface @LangChainAI @LlamaIndex @vLLMProject @RouteLLM @togethercompute @fireworks_ai @GroqInc @mistralai

Thanks. More deep dives into Leanstral and other AI models coming soon.

wow congrats! share more on leanstral

Que @MistralAI se démerde . C'est quoi ces bandes d'enfants gâtés qui ne peuvent se débrouiller seuls? Vous l'avez voulu ce "grand" monde occidental non ? Seulement voilà ça pêche du côté coqs politocards qui veulent leur poulailler exclusif #france . On arrête le copinage ?

Thank you 😊

Congrats

@MistralAI « Une impression de feuille morte » Hi, I Phi acts fluxion

Excited to share that I've joined the @MistralAI Ambassador program . Looking forward to supporting the community, running events, and helping grow the opensource AI ecosystem. Always happy to contribute to great open source projects.