Theorem proving is also starting to compete on cost: Mistral open-sources Leanstral 1.5, about $4 per problem.

According to Beating monitoring, Mistral AI has open-sourced Leanstral 1.5, a model for Lean 4 formal proofs. The model has a total of 119 billion parameters, with about 6.5 billion active parameters, uses the Apache-2.0 license, and offers free API access. Official evaluations show that Leanstral 1.5 solved 587 out of 672 problems on PutnamBench; it reached 87% and 34% on the abstract algebra benchmarks FATE-H and FATE-X, respectively, setting a best performance record among similar models. The average cost per problem solved by Leanstral 1.5 on PutnamBench is about $4, lower than the tens to hundreds of dollars cost of some earlier systems. As the per-problem token budget increases, the number of problems it solves continues to grow; in the AVL tree complexity proof, the model completed the related proof after more than 2.7 million tokens of reasoning and 22 context compressions. In addition to mathematical proofs, Leanstral 1.5 is also used for code verification. The team found 11 real bugs in 57 open-source Rust repositories, 5 of which had not been previously reported.
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pinned