Kimi K2.7 Code — Benchmarks

Benchmark scores for Kimi K2.7 Code aggregated from public leaderboards, with how it ranks among open models. See hardware requirements for what you need to run it.

Overall rank: #32 of 78 open modelscomposite 51.4/100 across 13 benchmarks in 4 categories · methodology

Coding

BenchmarkScoreOpen rankAll models
LiveBench Coding74.0#11 / 19#48 / 63
SciCode47.4%#13 / 62#79 / 161
WeirdML54.1%#8 / 46#60 / 161
LMArena WebDev1472.3#16 / 48#52 / 123
FrontierCode30.1%#4 / 11#22 / 36
DeepSWE30.5%#5 / 5#61 / 68
Surface Evolver Bench48.8%#5 / 13#18 / 27

Knowledge

BenchmarkScoreOpen rankAll models
SimpleQA36.5%#7 / 15#43 / 80

Math

BenchmarkScoreOpen rankAll models
FrontierMath54.0%#8 / 15#39 / 102
FrontierMath Tier 412.2%#9 / 11#52 / 64
LiveBench Math79.6#18 / 19#58 / 63
AIME 2024/202595.6%#5 / 82#37 / 270

Reasoning

BenchmarkScoreOpen rankAll models
Chess Puzzles21.0%#7 / 65#78 / 205
GPQA Diamond87.9%#13 / 94#57 / 289
SimpleBench57.9%#4 / 25#37 / 101
CritPt10.0%#9 / 62#69 / 172
LiveBench Reasoning82.8#9 / 19#39 / 63
APEX-Agents37.6%#7 / 10#27 / 33

Scores aggregated from public benchmark sources (each linked from the benchmark pages). llmrun does not run these benchmarks.