Kimi K3
Kimi K3 is an open-weights AI model from Moonshot with an index score of 168 (GPT-4 = 100), rank 30 of 217.
168152–183Index score#30 of 217 models
$0.19/ $16Price per million tokensinput / output · via OpenRouter
1.05MContext windowtokens it reads at once
16 Jul 2026Released4 months ago
Where it sitson the frontier
GPT-5.5 reached this score 3 months earlier.
Test resultsby area · tick = best by any model
Reasoning & knowledgeevidence
Mathsevidence
Codingevidence
Agentsevidence
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 9 leaderboards
The Model Indexthird-party tests, our method#30of 217 · 168
LMArena · Codingcoding questions in chat#12of 357 · 1538LMArena · Mathsmaths questions in chat#12of 346 · 1502LMArena · Instruction followingfollowing detailed instructions#14of 362 · 1489LMArena · Long promptslong, complex requests#14of 340 · 1500LMArena · Textoverall chat quality, judged by people#17of 362 · 1487LMArena · Creative writingcreative writing#25of 360 · 1459Epoch Capabilities Indexoverall capability from many benchmarks#25of 231 · 158.6LMArena · WebDevbuilding web apps and front ends#18of 133 · 1658Best attop three among current models
Not in the top three among current models on any category we track.
Where to use itprices checked 18 Nov 2026
API providers 8 of 19 · per 1M tokens, in / out
Run it yourself
Official weightsmoonshotai/Kimi-K3↗Host prices from OpenRouter; going direct can differ.
Facts and sources
Released16 Jul 2026
Announcementkimi.com ↗
Release entryKimi K3 (weights 07-27)
Scored asKimi K3
Output speed–
Hugging Facemoonshotai/Kimi-K3 ↗