Benchfolio

Leaderboard / Model

Kimi K3

Moonshot · Open weights · Released Jul 16, 2026 · $3 in · $15 out per 1M tokens

64Overall · rank 8

Score by job

Coding

77 · rank 7 · 3 of 5 benchmarks

BenchmarkResultVariantz
DeepSWE68.5%kimi-k3_max0.73
FrontierCode44.2%kimi-k3_unknown0.48
LMArena Coding1507 ratingkimi-k3-max0.80

Agents

Not enough data · 1 of 4 benchmarks

BenchmarkResultVariantz
APEX-Agents50.6%kimi-k3_unknown-0.16

Reasoning

49 · rank 19 · 3 of 4 benchmarks

BenchmarkResultVariantz
GPQA Diamond93.1%kimi-k3_max0.63
ARC-AGI-260.4%kimi-k3_max-0.16
SimpleBench60.7%kimi-k3_max-0.47

Math

61 · rank 11 · 5 of 5 benchmarks

BenchmarkResultVariantz
FrontierMath Tiers 1–372.2%kimi-k3_max0.29
FrontierMath Tier 439.0%kimi-k3_max-0.08
OTIS Mock AIME97.2%kimi-k3_max0.41
ProofBench87.0%kimi-k3_unknown1.55
LMArena Math1494 ratingkimi-k3-max0.51

Writing

68 · rank 12 · 2 of 2 benchmarks

BenchmarkResultVariantz
LMArena Creative Writing1452 ratingkimi-k3-max0.02
LMArena Instruction Following1478 ratingkimi-k3-max0.52

Design

64 · rank 7 · 1 of 1 benchmarks

BenchmarkResultVariantz
LMArena WebDev1660 ratingkimi-k3-max1.01