LongCat-Flash-Thinking
Announced on meituan.com
Benchmark results
| Benchmark | Score | Source check |
|---|---|---|
| AIME 2024No tools · average of 32 runs | 93.3% | CorroboratedThe same number is printed in 1 other lab document: 1 |
| AIME 2025No tools · average of 32 runs | 90.6% | |
| MMLU-Pro | 82.6% | |
| LiveCodeBenchaverage of 4 runs; problems from Aug 2024 to May 2025 | 79.4% | CorroboratedThe same number is printed in 1 other lab document: 1 |
| τ²-benchTelecom · average of 4 runs | 83.1% | |
| τ²-benchRetail · average of 4 runs | 71.5% | |
| τ²-benchAirline · average of 4 runs | 67.5% |
Sources
Reported by Meituan
- LongCat-Flash-Thinking model card (Hugging Face)huggingface.co · 22 Sep 2025
- LongCat-Flash-Thinking 正式发布，更强、更专业，保持极速！tech.meituan.com · 22 Sep 2025
Original image
