Signal
Mistral AI releases Leanstral 1.5, an interactive code agent advancing formal verification and bug detection
Evidence first: scan the strongest sources, then decide whether to go deeper.
Published 2026-07-03 22:32 UTCUpdated 2026-07-04 07:12 UTC
rsstelegram
modelsbenchmarkstooling
Source links open
Source links and full evidence are open here. Pro adds archive history, compare-over-time, alerts, exports, and workflow. Business adds Feed API integrations and team usage.
No card needed for the free brief.
Evidence trail (top sources)
top sources (1 domains)domains are deduped. counts indicate coverage, not truth.1 top source shown
limited source diversity in top sources
Overview
Mistral AI has launched Leanstral 1.5, a 119B parameter mixture-of-experts model designed as an interactive code agent integrated with the Lean 4 compiler.
Why now
- Released recently with open-source availability, enabling community adoption and further research.
- Achieves state-of-the-art results on challenging formal math benchmarks at reduced computational cost.
- Demonstrates the growing role of AI in formal methods and real-world code quality assurance.
Why it matters
- Introduces an interactive proof agent approach that improves formal verification accuracy and efficiency.
- Bridges formal theorem proving with practical software bug detection, enhancing code reliability.
- Offers an open-source tool advancing AI capabilities in mathematical reasoning and software engineering.
Evidence assessment
Recurring claims
- Leanstral 1.5 iteratively proposes proofs, reads compiler feedback, and refines output until successful compilation or resource limits are reached.
- Leanstral 1.5 solves 587 of 672 PutnamBench problems at about $4 per problem, significantly cheaper than comparable systems.
- The model found five previously unknown bugs while scanning 57 open-source repositories.
How sources frame it
- Machinelearningresearchnews: supportive
- Matthias Bastian: supportive
This release highlights a novel interactive approach to formal verification using AI, combining proof generation with compiler feedback for improved accuracy and efficiency.
All evidence
All evidence
Mistral's open-source Leanstral 1.5 aces formal math benchmarks and catches real bugs in code
The Decoder · the-decoder.com · 2026-07-04 07:12 UTC
Mistral AI Releases Leanstral 1.5: An Apache-2.0 Lean 4 Code Agent Model Solving 587 of 672 PutnamBench Problems
Marktechpost · marktechpost.com · 2026-07-03 22:32 UTC
Show filters & breakdown
Evidence items loaded: 0Publishers: 2Origin domains: 2Duplicates: -
Showing 2 / 2