Frontier Log

Meituan

LongCat-Flash-Thinking-2601

Announced on meituan.com

Benchmark results

BenchmarkScoreSource check
SWE-bench Verifiedaverage of 5 runs70.0%
AIME 2025With tools · with python; average of 16 runs99.6%
CorroboratedThe same number is printed in 1 other lab document: 1
heavy thinking mode; with python; average of 16 runs · parallel test-time compute100.0%
CorroboratedThe same number is printed in 1 other lab document: 1
LiveCodeBenchaverage of 4 runs; problems from Aug 2024 to May 202582.8%
CorroboratedThe same number is printed in 1 other lab document: 1
τ²-benchTelecom · average of 4 runs99.3%
τ²-benchRetail · average of 4 runs88.6%
τ²-benchAirline · average of 4 runs; airline environment fixes from the Claude Opus 4.5 report76.5%
Sources

Reported by Meituan

Related events

All models