Pass IndexThe State of AISign in

Kimi K2 Thinking

Kimi K2 Thinking is an advanced open-source thinking model by Moonshot AI.

text → text · made by Moonshot AI

$0.47$2per Mtok in / out
DeepInfra · 15 sellers

Sold by 16 ways

SellerLaneRate
AWS Bedrock cloudstandard$0.6per Mtok in$2.5per Mtok outmodels.dev · read 2026-09-16
Google Vertex AI cloudstandard$0.6per Mtok in$2.5per Mtok out$0.06per Mtok cachedcloud.google.com · read 2026-08-24
Microsoft Azure AI cloudstandard$0.6per Mtok in$2.5per Mtok outprices.azure.com · read 2026-08-24
Crusoe cloudstandard$2.5per Mtok in$2.5per Mtok outraw.githubusercontent.com · read 2026-09-16
DeepInfra aggregatorstandard$0.47per Mtok in$2per Mtok outapi.deepinfra.com · read 2026-08-25
Vercel AI Gateway aggregatorstandard$0.47per Mtok in$2per Mtok out$0.14per Mtok cachedai-gateway.vercel.sh · read 2026-08-25
IO.NET aggregatorstandard$0.55per Mtok in$2.25per Mtok out$0.28per Mtok cached$1.1per Mtok cache writemodels.dev · read 2026-09-16
Baseten aggregatorstandard$0.6per Mtok in$2.5per Mtok outraw.githubusercontent.com · read 2026-09-16
ElectronHub aggregatorstandard$0.6per Mtok in$2.5per Mtok outapi.electronhub.ai · read 2026-09-16
Fireworks AI aggregatorstandard$0.6per Mtok in$2.5per Mtok outraw.githubusercontent.com · read 2026-09-16
Hugging Face Inference Providers aggregatorstandard$0.6per Mtok in$2.5per Mtok out$0.15per Mtok cachedmodels.dev · read 2026-09-16
Nous Research aggregatorstandard$0.6per Mtok in$2.5per Mtok out$0.12per Mtok cachedinference-api.nousresearch.com · read 2026-09-12
Novita AI aggregatorstandard$0.6per Mtok in$2.5per Mtok out$0.15per Mtok cachedapi.novita.ai · read 2026-08-25
OpenRouter 2 ways$0.6per Mtok in$2.5per Mtok out$0.15per Mtok cached
OpenRouter aggregatorstandard$0.6per Mtok in$2.5per Mtok out$0.15per Mtok cachedopenrouter.ai · read 2026-08-24
OpenRouter aggregatorgoogle-vertex$0.6per Mtok in$2.5per Mtok outopenrouter.ai · read 2026-08-25
PPIO aggregatorstandard$4per Mtok in$16per Mtok outapi.ppinfra.com · read 2026-09-16

Measured 13 standings

PlaceBoardMetricScore
13thof 61AIME 2025Accuracy (± 95% CI)92.5
15thof 23Cl bench — Epoch AIOverall0.176
23rdof 47SWE-bench Verified (default "Bash Only" view, agent = mini-SWE-agent)% Resolved63.4
25thof 50Metr time horizons — Epoch AIaverage_score0.592
37thof 101FrontierMath (Tiers 1-3)mean_score21.404
60thof 65Apex agents — Epoch AIPass@1 score0.041
65thof 72Frontiermath tier 4 — Epoch AImean_score0
84thof 112Ale bench — Epoch AIPerformance597.5
98thof 170Weirdml — Epoch AIAccuracy42.79
114thof 202Terminalbench — Epoch AIAccuracy mean0.357
122ndof 162Design Arena · website (models)Elo1125
148thof 184Artificial Analysis · Coding IndexIndex21
309thof 656Epoch capabilities index — Epoch AIECI Score145.79

About

Kimi K2 Thinking — a text model from Moonshot AI, sold by 15 companies from $0.47 in and $2 out per million tokens, placed 13th of 61 on AIME 2025.

It takes text and returns text, with a context window of 216,144 tokens. It was published in November 2024, trained on material up to August 2024. Its sellers say it can reason step by step and call a tool. The catalogue files it under reasoning, code and agents. Fifteen companies sell it. The cheapest is $0.47 in and $2 out per million tokens at DeepInfra. Beside the standard rate there is a separately routed lane. It has been measured on 13 boards, and stands best at 13th of 61 on AIME 2025.

Every current figure

Maker
Moonshot AI
Register
model
Takes
text
Returns
text
Context
216,144 tokens
Published
November 2024
Knowledge to
August 2024
Parameters
1058 billion
Licence
not read
Sellers
15
Maker's own price
not read
Price
$0.47 in and $2 out per million tokens — DeepInfra
Boards
13
Best place
13th of 61 — AIME 2025

Known as 3 names

K2 Thinkingmoonshotai/Kimi-K2-ThinkingKimi-K2-Thinking