AIToday
Large Language ModelsOpen-Source AIAI Business & IndustryTHE DECODERPublished: Jul 4, 2026, 19:00 JST1 min read

Mistral's Leanstral 1.5 hits 100% on formal math benchmark, spots real code bugs

Mistral's Leanstral 1.5 hits 100% on formal math benchmark, spots real code bugs

Key takeaway

  • Mistral AI released Leanstral 1.5, an open-source model trained to formally verify mathematical proofs and software correctness in the Lean 4 language.

  • It achieves perfect scores on some math benchmarks and ranks at the top of open-source models on several formal algebra tests, and in real-world testing it identified five previously unknown bugs in actual open-source code repositories.

3 Key Points

  1. What happened

    Mistral AI released Leanstral 1.5, a free open-source model licensed under Apache 2.0, designed for formal verification in Lean 4 (a programming language for verifying math proofs and software correctness). The model scores 100 percent on miniF2F, solves 587 of 672 problems on PutnamBench, and achieves 87 and 34 percent on the algebra benchmarks FATE-H and FATE-X. In practical testing, it scanned 57 open-source repositories and caught five previously unknown bugs, including an overflow bug in the Rust library varinteger.

  2. Why it matters

    The model ranks at the top of the open-source field on PutnamBench, FATE-H, and FATE-X—only the closed-source Aleph Prover surpasses it on PutnamBench. For software teams and mathematicians, this suggests that open-source formal verification is reaching a level where it can detect real-world defects in production code, potentially reducing costly bugs before deployment.

  3. What to watch

    Leanstral 1.5 is available now through Hugging Face and via a free API, making it immediately accessible to developers and researchers.

Ask the AI about this article →

FAQ

Is there a cost to use Leanstral 1.5?
No. The model is free and licensed under Apache 2.0, and is available through Hugging Face and via a free API.
What is Leanstral 1.5 designed to do?
It is built for formal verification in the Lean 4 programming language, which is designed to formally verify mathematical proofs and software correctness. Although trained mainly for math, Mistral says it also performs well at code verification.

Get the latest Large Language Models news every morning

For example, today's edition would include:

  • Visko raises $10M, launches live AI video model OrbisSiliconANGLE AI · 1h ago
  • Runway unveils Solaris, an AI that generates app interfaces in real timeTHE DECODER · 1h ago
  • Google AI Search flags Facebook users as dangerTHE DECODER · 1h ago

AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.

Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. Q&As are published on this page for other readers too.

Related Articles

Next articleTexas oilman funds advanced nuclear startup at Christian college