Skip to content
Yardstick
  1. Models
  2. Moonshot AI
  3. Kimi K3

Kimi K3

Moonshot AI · Flagship · Jul 16, 2026Open weights

Moonshot AI's flagship: openly published weights, a one-million-token context and a strong showing on hard software-engineering tasks.

Announcement (opens in a new tab)
Artificial Analysis Intelligence Index
43.6#12 of 29 · max effort
API price per 1M tokens
$3 / $15input / output · $6 blended
Context window
1Mtokens
Output speed
42 t/s51s to first answer at max effort

Benchmark results

Each row shows this model (blue) against every other model in the catalogue (grey), worst on the left, best on the right.

Overall capability

  • Artificial Analysis Intelligence Index43.6#12 of 29

    Max effort

    11.357.6
  • Epoch Capabilities Index157.4#6 of 16
    146.9167.3
  • LiveBench79.2%#7 of 23
    63.9%83.4%

Coding

  • SWE-Bench Pro (V2 Hard)88.2%#3 of 5

    Mini-swe-agent harness, max effort

    58.8%92.2%
  • Arena WebDev1,655#9 of 27

    Max effort

    1,2631,814

Agents & professional work

  • Vending-Bench 2$5,165#7 of 9
    $2,158$15,515

Reasoning

  • ARC-AGI-260.4%#9 of 11

    Max effort

    42.4%95%
  • FrontierMath (Tiers 1–3)72.2%#9 of 14

    Max effort

    53%93.7%
  • MathArena50.5%#6 of 10
    35.3%90.1%
  • SimpleBench60.7%#11 of 12

    Max effort

    45.8%88.4%

Knowledge & science

  • Humanity's Last Exam (Diamond)22.2%#7 of 9
    13.4%60.6%
  • GPQA Diamond93.1%#6 of 14

    Max effort

    90.2%95.8%

Human preference

  • Arena Text1,488#6 of 24

    Max effort

    1,4251,525

Reasoning effort trade-offs

The same model at different thinking budgets. More effort buys a higher intelligence score at a higher cost per task and a longer wait. Measured by Artificial Analysis.

EffortScoreCost / taskFirst answerSpeed
Low
30.1
$1.1553s41 t/s
Max (headline score)
43.6
$2.0051s42 t/s

Specs & pricing

Developer
Moonshot AI
Released
Jul 16, 2026
Availability
Generally available
API model ID
kimi-k3
Weights
Open · Kimi K3 License
Parameters
2.8T (MoE)
Reasoning
Always on
Accepts
Text, Image, Video
Produces
Text
Context window
1,048,576 tokens
Input price
$3 / 1M tokens
Output price
$15 / 1M tokens
Cached input
$0.30 / 1M tokens

Closest alternatives

Models from other labs nearest to Kimi K3 on Artificial Analysis Intelligence Index.

  • Claude Haiku 5.5

    Anthropic

    Intelligence 43.4 · $0.10 / $0.50

    Compare side by side
  • GLM-5.3

    Z.ai

    Intelligence 44.8 · $1.40 / $4.40

    Compare side by side
  • GLM-5.3-Flash

    Z.ai

    Intelligence 41.8 · $0.15 / $0.50

    Compare side by side

Sources

  • Announcement (opens in a new tab)
  • Model documentation (opens in a new tab)
  • Pricing (opens in a new tab)
Yardstick

A plain reading of how today’s SI models compare. Scores are taken from each benchmark’s own leaderboard wherever one exists; anything a lab reported about its own model is marked.

Data as of Oct 8, 2026.

Explore

  • Models
  • Compare
  • Benchmarks
  • Value
  • Labs
  • Methodology & sources

Sources

  • Artificial Analysis (opens in a new tab)
  • Arena (opens in a new tab)
  • Epoch AI Benchmarking Hub (opens in a new tab)
  • ARC Prize (opens in a new tab)
  • Terminal-Bench (opens in a new tab)
  • Scale AI Labs (opens in a new tab)

Yardstick is an independent reference and is not affiliated with, sponsored by or endorsed by any lab listed. Product names and logos are trademarks of their respective owners and are used only to identify them. Benchmarks are imperfect proxies: check a model on your own task before you commit to it.

  • Models
  • Compare
  • Benchmarks
  • Value
  • Labs
  • Models
  • Compare
  • Benchmarks
  • Value
  • Labs
  • Models
  • Compare
  • Benchmarks
  • Value
  • Labs