Leanstral 1.5

Mistral has unveiled Leanstral 1.5, a new AI model built to help with formal math-style proofs and turning human-written statements into machine-checkable logic. In plain English: it’s meant for the kind of ultra-picky reasoning work where computers verify every step. It’s big, it’s specialized, and it comes with a huge context window and a very attention-grabbing listed price of $0 — but the real fireworks were in the comments.

Almost immediately, the crowd split into camps. One side asked the brutal question hanging over the whole launch: is anyone using Mistral because it’s actually the best, or just “because EU”? That line from doctorpangloss landed like a match in dry grass, turning the thread into a low-key identity crisis about whether Mistral wins on merit or on European loyalty. Meanwhile, others were less interested in geopolitics and more interested in usefulness: could this thing help specify real software, or is it only for pure theorem nerds?

Then came the classic specialist-model drama. One commenter raised an eyebrow that Leanstral is tuned for Lean 4 but not rival proof systems like Coq, which is exactly the kind of niche academic shade the internet lives for. And in a very startup-meets-open-source twist, another commenter popped in to say they’d just launched OpenATP, a tool that already supports Leanstral and now needs a quick update because the old version was deprecated. So yes: new model, old model already out, and commenters asking whether this is a breakthrough, a branding play, or just another very smart AI looking for a clear audience.