Mistral AI has released Leanstral 1.5 as an open-source model designed to verify mathematical proofs and check whether software behaves correctly.

The company published the model under the permissive Apache 2.0 licence, allowing developers and businesses to use, modify and deploy it commercially. Organisations can also self-host Leanstral 1.5, which may help those with strict privacy, compliance, and data-governance requirements.

Mistral Leanstral 1.5 is designed to work with Lean 4, an interactive theorem prover used to express and verify mathematical statements and software specifications.

The model targets academic mathematics as well as practical software verification, where developers need to prove that code meets defined correctness and safety requirements.

Mistral is offering Leanstral 1.5 through a free API endpoint. Developers can also download it from Hugging Face for self-hosted deployments.

Leanstral 1.5 uses a Mixture-of-Experts architecture with 119 billion total parameters.

Only around 6 billion parameters are active during inference, allowing the model to use a smaller portion of its full architecture while processing each request.

According to Mistral, Leanstral 1.5 achieved a perfect score on both the validation and test portions of miniF2F, a benchmark used to evaluate formal mathematical reasoning.

The model also solved 587 of the 672 problems included in PutnamBench, which tests complex reasoning and lengthy mathematical proofs.

Mistral said Leanstral required approximately one-seventh of the computing resources that Claude Opus 4.6 would have consumed while attempting the same PutnamBench workload.

Leanstral 1.5 also solved 87% of the graduate-level abstract algebra problems included in FATE-H and 34% of the doctoral-level problems in FATE-X.

Mistral also demonstrated a system that uses Leanstral 1.5 to verify Rust software automatically.

The pipeline uses Aeneas to translate Rust code into Lean representations. Leanstral can then attempt to prove whether the code satisfies its defined correctness properties or identify cases where those properties are violated.

Mistral tested the system across 57 open-source Rust repositories. The pipeline detected 47 violated properties and confirmed 11 genuine software bugs.

Five of those bugs had not previously been reported through GitHub, according to the company.

The Apache 2.0 licence allows organisations to integrate Leanstral 1.5 into commercial products, modify the model, and deploy it on their own infrastructure.

Self-hosting also allows companies to keep sensitive source code, mathematical research, and verification data within their own systems rather than sending it to a third-party hosted service.

The release follows Mistral’s introduction of Mistral OCR 4, its latest document-intelligence and optical-character-recognition model.

Get the latest tech news, telecom insights, and product launches wherever you prefer.

Add ProPakistani to Preferred Sources and see more of our stories in Google Search and Top Stories.

Shares