Kimi K2.5
Kimi K2.5 is an open-weights AI model from Moonshot with an index score of 145 (GPT-4 = 100), rank 79 of 217.
145130–161Index score#79 of 217 models
$0.50/ $2.5Price per million tokensinput / output · via OpenRouter
262KContext windowtokens it reads at once
27 Jan 2026ReleasedReplaced by Kimi K2.6
Where it sitson the frontier
o3 reached this score 9 months earlier.
Test resultsby area · tick = best by any model
Reasoning & knowledgeevidence
Mathsevidence
Codingevidence
Agentsno evidence yet
Professional workevidence
Novel problemsevidence
Long tasksno evidence yet
Third-party results only, at the model’s best standard setting. Hover a bar for who ran it.
What others sayrank on 11 leaderboards
The Model Indexthird-party tests, our method#79of 217 · 145
LMArena · Mathsmaths questions in chat#49of 346 · 1470LMArena · Codingcoding questions in chat#62of 357 · 1505LMArena · Textoverall chat quality, judged by people#69of 362 · 1450LMArena · Instruction followingfollowing detailed instructions#69of 362 · 1439LMArena · Creative writingcreative writing#69of 360 · 1422LMArena · Long promptslong, complex requests#69of 340 · 1457LMArena · Visionunderstanding images#43of 145 · 1252Epoch Capabilities Indexoverall capability from many benchmarks#74of 231 · 148.8LMArena · WebDevbuilding web apps and front ends#77of 133 · 1437LMArena · Documentsworking with long documents#32of 40 · 1431Best attop three among current models
Where to use itprices checked 18 Nov 2026
API providers 4 · per 1M tokens, in / out
Run it yourself
Official weightsmoonshotai/Kimi-K2.5↗Host prices from OpenRouter; going direct can differ.
Facts and sources
Released27 Jan 2026
Announcementkimi.com ↗
Release entryKimi K2.5
Scored asKimi K2.5
Output speed–
Hugging Facemoonshotai/Kimi-K2.5 ↗