Back

Moonshot AI

Kimi K2.6

ReasoningOpen weightsModels.devHugging Face
Intelligence
27.0
Rank 152 of 744 measured models
Coding
61.8
Rank 79 of 298 measured models
Output speed
52.8 t/s
Rank 307 of 403 measured models
Price/1M
$1.71
Rank 451 of 659 measured models
Context window
256K
Rank 518 of 1199 measured models
Cost per task
$0.80
Rank 118 of 193 measured models

Kimi K2.6 (Reasoning) scores 27.0 on the Artificial Analysis Intelligence Index, ranking 152nd of 744 measured models. The closest models on this index are Grok Build 0.1 0616 (27.2), Qwen3.6 Plus (27.0), and Agnes 2.5 Pro Alpha (based on Qwen3.5 397B A17B) (26.8). At $1.71 per million tokens (3:1 input/output blended price), it is more expensive than Grok Build 0.1 0616 ($1.25), Qwen3.6 Plus ($1.13), and Agnes 2.5 Pro Alpha (based on Qwen3.5 397B A17B) ($0.563). With an output speed of 52.8 t/s, it is slower than Grok Build 0.1 0616 (67.4 t/s), Qwen3.6 Plus (55.9 t/s), and Agnes 2.5 Pro Alpha (based on Qwen3.5 397B A17B) (160 t/s).

Artificial Analysis Indices
Reasoning
Intelligence27.0
Rank 152 of 744 measured models
Coding61.8
Rank 79 of 298 measured models
Agentic20.5
Rank 97 of 189 measured models
Reasoning levels
Each point is a reasoning level of this model: more reasoning often raises the score but costs more. Select a point to open that level.
Intelligence Indexmost efficient ↗Price / 1M tokens
Capability indexes
Performance in specific industries, measured by Artificial Analysis
Finance & Accounting28.3
Rank 122 of 218 measured models
Strategy & Ops22.0
Rank 138 of 218 measured models
Legal29.4
Rank 123 of 218 measured models
Engineering30.5
Rank 112 of 217 measured models
Economics40.5
Rank 106 of 225 measured models
Standard Benchmarks
GPQA Diamond91.1%
Rank 52 of 619 measured models
HLE37.5%
Rank 102 of 664 measured models
SciCode51.5%
Rank 94 of 231 measured models
IFBench76.0%
Rank 24 of 450 measured models
AA-LCR81.0%
Rank 59 of 575 measured models
TerminalBench Hard43.9%
Rank 37 of 432 measured models
Terminal-Bench 2.165.9%
Rank 83 of 244 measured models
Terminal-Bench v40.5%
Rank 136 of 218 measured models
Additional benchmarks21
Design Arena · 3d57.6%
Design Arena · Agenticgamedev47.5%
Design Arena · Agentichtmlslides59.0%
Design Arena · Agenticslides Html59.2%
Design Arena · Agenticslides Python Pptx45.5%
Design Arena · Agenticslides45.8%
Design Arena · Androidnative49.1%
Design Arena · Asciiart47.5%
Design Arena · Codecategories54.8%
Design Arena · Dataviz52.0%
Design Arena · Fullstack49.4%
Design Arena · Gamedev54.9%
Design Arena · Godotgamedev46.8%
Design Arena · Htmlslides52.3%
Design Arena · Mobileapps47.6%
Design Arena · Pptxslides44.3%
Design Arena · Python Pptxslides42.1%
Design Arena · Svg51.2%
Design Arena · Uicomponent55.6%
Design Arena · Webapps59.3%
Design Arena · Website54.2%
Models at a similar level
The closest models on the Intelligence Index, one configuration per family.

Preferences and analytics

Rybbit: visits, performance, clicks, copies, forms and session replay (masked inputs), with your permission. Refusing does not limit the site.

Privacy

Allow navigation analytics?

With your permission, Rybbit measures visits, performance and interactions (clicks, copies, forms), including session replay, to improve BenchSift. Input values are masked in replays. Refusing does not limit the site.

Your choice is stored for 180 days. Change it at any time using “Privacy preferences” in the footer. Withdrawal reloads the page.

No choice stored: Rybbit is disabled.

Privacy