Dark Side of the Moon Releases k0-math Reasoning Model, Mathematical Ability to Benchmark OpenAI o1 Series

On the one-year anniversary of the opening of Kimi Chat to the entire community, Dark Side of the Moon announced the launch of its new first-generation mathematical reasoning model, k0-math. benchmark tests show that the mathematical capabilities of Kimi k0-math can be benchmarked against the world's two publicly-available models of the world's leading OpenAI o1 series: o1-mini and o1-preview. in four benchmark tests, including the middle school exam, the college entrance exam, and MATH, which contains introductory competition questions, the k0-math first-generation model outperforms the o1-mini and o1-preview models, Exam, and MATH, which contains introductory competition questions, the k0-math primitive model outperforms the o1-mini and o1-preview models in four math benchmark tests. In the OMNI-MATH and AIME benchmarks, which are more difficult competition-level math problems, the k0-math primitive model achieves the highest performance of 90% and 83% of o1-mini, respectively.(Sina Finance)