Mistral AI Open-Sources Leanstral 1.5 to Lower Barriers in Mathematical Research
Mistral AI has recently released an open-source model named Leanstral1.5, built specifically for the Lean4 mathematical formal proof language. Licensed under Apache-2.0, this model features 119B total parameters with only 6B activated, keeping performance high while significantly cutting costs.
As a specialized tool for mathematical reasoning, Leanstral1.5 delivers impressive results. On the respected miniF2F formal mathematics benchmark, it achieved a 100% completion rate on both the validation and test sets. When tackling the challenging PutnamBench competition problems, the model successfully solved 587 out of 672 Lean4 questions. It also excelled in the FATE benchmark series for abstract algebra, posting an 87% success rate on the master-level FATE-H test and a 34% success rate on the doctoral-level FATE-X test—setting a new record for models of this kind.

Cost efficiency stands out as another key advantage of this release. Mistral AI highlights that compared to existing alternatives, Leanstral1.5 dramatically lowers the expense of scientific trial and error. For instance, solving a question from PutnamBench with Leanstral1.5 costs an average of just $4, whereas the comparative model Seed-Prover1.5 costs over $300, and Aleph Prover ranges from $54 to $68. This sharp reduction in cost is expected to bring high-precision mathematical proof assistance out of the lab and into broader research use.
In practical code development scenarios, Leanstral1.5 also demonstrates strong bug-finding ability. When tested on 57 code repositories, it identified 47 violations, 11 of which were confirmed as genuine defects. Notably, five of those vulnerabilities had never been reported on GitHub before, underscoring the model’s potential to assist in program verification and security audits.
With the open-source release of Leanstral1.5, the fields of mathematics and computer science gain easier access to a powerful proof-assistance tool. By reducing both computational and financial costs, this model is poised to accelerate the adoption of formal mathematical proofs, helping researchers move beyond tedious calculations and verification to focus on fundamental scientific breakthroughs.
Related article
DeepMind CEO Hassabis: I sleep six hours a day, usually feel energetic around 1 a.m.
Fortune recently featured an interview with Demis Hassabis, CEO of Google DeepMind, revealing his unconventional approach to rest and productivity. Hassabis disclosed that he sleeps very little, structuring his waking hours into two distinct work blo
OpenAI, Anthropic Vie for Market Share Despite Revenue Shortfalls
Despite recent reports suggesting OpenAI missed revenue targets, creating pressure on tech stocks this Tuesday, private AI lab investors remain resilient. Seasoned backers have confirmed they will not reduce investment despite negative media coverage
California AV Compliance: A New Era of Tickets, Geofences, and 1M Miles
Guident operates an AuveTech shuttle in South Florida, managing a four-mile route in West Palm Beach and a one-mile route in Boca Raton using its remote monitoring technology. | Credit: GuidentCalifornia is redefining the regulatory landscape for dri
Related Special Topic Recommendations
Comments (0)
0/500
Mistral AI has recently released an open-source model named Leanstral1.5, built specifically for the Lean4 mathematical formal proof language. Licensed under Apache-2.0, this model features 119B total parameters with only 6B activated, keeping performance high while significantly cutting costs.
As a specialized tool for mathematical reasoning, Leanstral1.5 delivers impressive results. On the respected miniF2F formal mathematics benchmark, it achieved a 100% completion rate on both the validation and test sets. When tackling the challenging PutnamBench competition problems, the model successfully solved 587 out of 672 Lean4 questions. It also excelled in the FATE benchmark series for abstract algebra, posting an 87% success rate on the master-level FATE-H test and a 34% success rate on the doctoral-level FATE-X test—setting a new record for models of this kind.

Cost efficiency stands out as another key advantage of this release. Mistral AI highlights that compared to existing alternatives, Leanstral1.5 dramatically lowers the expense of scientific trial and error. For instance, solving a question from PutnamBench with Leanstral1.5 costs an average of just $4, whereas the comparative model Seed-Prover1.5 costs over $300, and Aleph Prover ranges from $54 to $68. This sharp reduction in cost is expected to bring high-precision mathematical proof assistance out of the lab and into broader research use.
In practical code development scenarios, Leanstral1.5 also demonstrates strong bug-finding ability. When tested on 57 code repositories, it identified 47 violations, 11 of which were confirmed as genuine defects. Notably, five of those vulnerabilities had never been reported on GitHub before, underscoring the model’s potential to assist in program verification and security audits.
With the open-source release of Leanstral1.5, the fields of mathematics and computer science gain easier access to a powerful proof-assistance tool. By reducing both computational and financial costs, this model is poised to accelerate the adoption of formal mathematical proofs, helping researchers move beyond tedious calculations and verification to focus on fundamental scientific breakthroughs.
DeepMind CEO Hassabis: I sleep six hours a day, usually feel energetic around 1 a.m.
Fortune recently featured an interview with Demis Hassabis, CEO of Google DeepMind, revealing his unconventional approach to rest and productivity. Hassabis disclosed that he sleeps very little, structuring his waking hours into two distinct work blo
OpenAI, Anthropic Vie for Market Share Despite Revenue Shortfalls
Despite recent reports suggesting OpenAI missed revenue targets, creating pressure on tech stocks this Tuesday, private AI lab investors remain resilient. Seasoned backers have confirmed they will not reduce investment despite negative media coverage





Home






