Support

Moonshot · released 2025-11-06

Kimi K2 Thinking

Measured on 95 benchmarks.

Compare

Not ranked in this one: too little measured data for its benchmarks. Pick another ranking above.

By category

  • Kimi K2 Thinking
  • Fable 5.2 (#1)
  • Field median
p25p50p75topField medianCodingAgents & computeruseProfessional workScience & researchMathReasoning &puzzlesGames & strategyKnowledge &accuracyLong documentsWriting & designBehavior & styleHuman preference
  1. Coding6 benchmarks5%#102 of 107
  2. Agents & computer use4 benchmarks23%#80 of 110
  3. Professional work5 benchmarks6%#102 of 110
  4. Science & research5 benchmarks9%#96 of 106
  5. Math5 benchmarks1%#57 of 59
  6. Reasoning & puzzles6 benchmarks9%#93 of 102
  7. Games & strategy2 benchmarks2%#66 of 66
  8. Knowledge & accuracy8 benchmarks20%#87 of 103
  9. Long documents3 benchmarks12%#91 of 102
  10. Writing & design4 benchmarks13%#61 of 64
  11. Behavior & style3 benchmarks54%#18 of 50
  12. Human preference2 benchmarks9%#62 of 66

Across 53 results on every current benchmark that measured Kimi K2 Thinking, not only this ranking’s. The percentage is its average standing among the 109 models in this ranking that took the same benchmarks: 100% beat them all, 0% trailed them all. A benchmark in two categories counts in both; a category needs two benchmarks.

Availability

API providers

3
ProviderIn / out per 1MTok/sFirst
  • Amazon Bedrock$0.60 / $2.501113.6s
  • Microsoft Azure$0.60 / $2.501352.9s
  • Novita$0.60 / $2.50423.9s

OpenRouter, Artificial Analysis

This ranking’s benchmarks

0 of 10 measured

Cost per task is what a subscriber pays: the model’s plan price, a share of API price. Models no plan sells are at API price.

This model has no results on this ranking’s benchmarks.

All 95 benchmarks

show 95 more