Kimi K2.5

MoonshotReleased 27 Jan 2026Open

Compare Announcement

149138–161Index score#60 of 229 models

5/ 8Core tests taken3 of 6 areas · plus 12 anchor tests

−13Versus the frontier at releasebest then: GPT-5.2Replaced by Kimi K2.6

$0.45/ $2.25Price per million tokensinput / output, list price

262KContext windowtokens it reads at once

Where it sitson the frontier

GPT-5 reached this score 6 months earlier.

Test resultsby area · tick = best by any model

Reasoning & knowledgeevidence

GPQA Diamond

88%

Humanity's Last Exam

24%

Mathsevidence

FrontierMath T1-3

no result

Codingevidence

SWE-bench Verified

74%

Terminal-Bench

43%

Agentsevidence

APEX-Agents

no result

Novel problemsevidence

ARC-AGI-2

12%

Long tasksno evidence yet

METR time horizon

no result

Independent results from the Epoch AI benchmarking hub, best reasoning setting. The core tests are shown here; anchor tests also inform the score and can be shown above.

What others sayrank on each leaderboard

The Model Indexindependent tests, our method#60of 219 · 149

LMArena · Mathsmaths questions in chat#45of 373 · 1471 LMArena · Codingcoding questions in chat#65of 383 · 1502 LMArena · Textoverall chat quality, judged by people#68of 388 · 1450 LMArena · Creative writingcreative writing#69of 386 · 1423 Epoch Capabilities Indexoverall capability from many benchmarks#58of 239 · 148.0 LMArena · Visionunderstanding images#42of 149 · 1252 LMArena · WebDevbuilding web apps and front ends#66of 123 · 1436 LMArena · Documentsworking with long documents#32of 41 · 1430

Further left is better. Leaderboards measure different things: LMArena is people voting blind between two answers, the others are test scores. Click any row for the source.