Benchmark Almanac

LongCat-2.5

LongCat-2.5 ranks first on 3 of the 22 benchmarks with five or more models scored.

Meituan Released November 2nd, 2026 Family LongCat The lab's model card Compare with others
28
Benchmarks with a score
3
Ranked first, out of five or more models
7
In the top three
Meituan Nov 2026

LongCat-2.5

Closed weights Flagship

163.8 #17 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
  • NL2Repo-Bench 60.5% #6 /30
  • Claw-Eval Pass^3 78.3% #1 /13
  • FinSearchComp 63.4% #5 /7

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
Humanity's Last Exam 63.1% #10 of 154 Lab model card

Math and science

BenchmarkScoreRankSource
FrontierScience 77% #2 of 8 Lab model card

Coding

BenchmarkScoreRankSource
PaperBench 82.1% #2 of 15 Lab model card
Vision2Web 78.6% #1 of 7 Lab model card
NL2Repo-Bench 60.5% #6 of 30 Lab model card
SWE Atlas 58.3% #6 of 14 Lab model card
DeepSWE 64.6% #32 of 56 Lab model card
ProgramBench almost solved 25.5% #6 of 8 Lab model card
SWE Atlas Refactoring 55.2% #1 of 1 Lab model card
SWE Atlas Test Writing 47.8% #1 of 1 Lab model card

Agents and tools

BenchmarkScoreRankSource
BrowseComp 91.9% #4 of 81 Lab model card
Claw-Eval Pass^3 78.3% #1 of 13 Lab model card
Terminal-Bench 2.1 88.4% #8 of 56 Lab model card
ResearchRubrics 81% #1 of 6 Lab model card
AutomationBench 51.8% #10 of 47 Lab model card
SkillsBench 63.3% #3 of 14 Lab model card
MCP Atlas 83.1% #12 of 46 Lab model card
Terminal-Bench 4.0 38.5% #18 of 28 Lab model card

Professional work

BenchmarkScoreRankSource
JobBench 65.4% #4 of 15 Lab model card
3DCodeBench 55.4% #1 of 3 Lab model card
SpreadsheetBench 2 34.9% #1 of 3 Lab model card
FORTE 83.3% #1 of 2 Lab model card
FinSearchComp 63.4% #5 of 7 Lab model card
Agents' Last Exam v1.1 38.5% #3 of 3 Lab model card

Vision, audio and robotics

BenchmarkScoreRankSource
MATH-Vision (MathVision) 96.5% #2 of 51 Lab model card
CharXiv 90.2% #8 of 75 Lab model card
Video-MME 85.2% #20 of 51 Lab model card

Other

BenchmarkScoreRankSource
SuperCLUE 39.6 #14 of 16 Official leaderboard

Its lab's numbers put it at 163.8, number 17 of 404

The Self-Reported Index fits one score to the 17 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 161.5 to 166.0. How we compute it.

Its launch post

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

Introducing LongCat-2.5 (LongCat blog) Nov 2, 2026