Open-source models can solve math competition problems to this extent, with each problem costing only $4. It feels like the pricing logic of closed-source commercial models needs to be rethought.

CoinWorld News: Mistral’s open-source Leanstral 1.5 model has an estimated problem-solving cost of about $4 per question, with a total parameter count of 119.0 billion and approximately 6.5 billion activated parameters. It uses the Apache-2.0 license and provides free API access. Official evaluations show that Leanstral 1.5 solved 587 out of 672 questions on Putnambench, reaching 87% and 34% on the abstract algebra benchmarks Fate-H and Fate-X, respectively, and setting the best performance among comparable models. In addition, Leanstral 1.5 has also been used for code verification, with the team finding 11 real bugs across 57 open-source Rust repositories, 5 of which had not been previously reported.