Benchmark Almanac

LongCat-Flash-Thinking

LongCat-Flash-Thinking ranks first on MATH-500.

Meituan Released September 21st, 2025 Family LongCat-Flash The lab's model card Compare with others
25
Benchmarks with a score
1
Ranked first, out of five or more models
2
In the top three
Meituan Sep 2025

LongCat-Flash-Thinking

Open weights Flagship

141.1 #168 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
  • AMO-Bench 43.6% #4 /21
  • General365 44.4% #22 /23
  • tau2-bench 74% #50 /79

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

BenchmarkScoreRankSource
MMLU-Pro 82.6% #55 of 167 Lab model card
IFEval 86.9% #64 of 135 Lab model card
Arena-Hard v2.0 69.9% #12 of 24 Lab model card
MMLU-Redux 89.3% #27 of 45 Lab model card
Extended NYT Connections 17.7% #88 of 103 Official leaderboard

Math and science

BenchmarkScoreRankSource
MATH-500 99.2% #1 of 70 Lab model card
AIME 2024 93.3% #6 of 88 Lab model card
ZebraLogic 95.5% #3 of 23 Lab model card
AMO-Bench 43.6% #4 of 21 Official leaderboard
AIME 2025 90.6% #47 of 143 Lab model card
GPQA Diamond 81.5% #107 of 302 Lab model card
Beyond AIME 69.5% #5 of 14 Lab model card
HMMT 2025 83.7% #34 of 54 Lab model card
miniF2F 81.6% #4 of 5 Lab model card
ARC-AGI 50.3% #20 of 21 Lab model card
General365 44.4% #22 of 23 Official leaderboard
Sudoku-Bench 56% #1 of 1 Lab model card

Coding

BenchmarkScoreRankSource
LiveCodeBench 79.4% #11 of 54 Lab model card
OJBench 40.7% #6 of 16 Lab model card
LiveCodeBench v6 79.4% #34 of 90 Lab model card
SWE-bench Verified 59.4% #113 of 155 Lab model card

Agents and tools

BenchmarkScoreRankSource
Berkeley Function Calling Leaderboard (BFCL) 74.4% #7 of 41 Lab model card
tau2-bench 74% #50 of 79 Lab model card
VITA-Bench 29.5% #13 of 16 Lab model card

Safety and security

BenchmarkScoreRankSource
LongCat safety evaluation (Meituan) 95.7% #1 of 2 Lab model card

Its lab's numbers put it at 141.1, number 168 of 404

The Self-Reported Index fits one score to the 19 capability results Meituan reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 139.3 to 142.8. How we compute it.

Its launch post

The benchmarks the lab chose to headline, with the scores it gave.

LongCat-Flash-Thinking model card (Hugging Face) Sep 21, 2025

Its system card

Every result in the longer documents the lab published with it. 1 of them isn't in the launch post.