in the loop

Kimi K2.5
#64 of 274 scored.

MoonshotJan 27, 2026Open weights

Capabilities Index

Range 146.4–149.3

148.0

Context

Takes text, image

262K

Input

Per 1M tokens

$0.45

Output

Per 1M tokens

$2.25

Benchmarks

How it scores.

  • GPQA Diamond83.5%

    53rd of 186

  • SWE-bench Verified73.8%

    17th of 32

  • ARC-AGI-211.8%

    43rd of 79

  • Humanity's Last Exam20.6%

    15th of 40

  • SimpleQA Verified34.3%

    51st of 77

  • Mock AIME92.2%

    46th of 176

  • Terminal-Bench43.2%

    18th of 35

Everything else Epoch has for it

  • Fiction.LiveBench86.1%
  • ARC-AGI65.3%
  • OSWorld63.3%
  • FrontierMath-2025-02-28-Private48.9%
  • WeirdML45.6%
  • SimpleBench36.2%
  • CL-bench19.3%
  • CL-bench Life13.2%
  • Chess Puzzles7.4%
  • FrontierMath-Tier-4-2025-07-01-Private7.0%

On the index

Around it.

  1. #61DeepSeek-V4-ProDeepSeek · $0.21 in149.1
  2. #62GPT-5.4 MiniOpenAI · $0.75 in148.8
  3. #63InklingThinking Machines · $1.00 in148.5
  4. #64Kimi K2.5Moonshot · $0.45 in148.0
  5. #65Qwen 3.6 PlusAlibaba · $0.33 in147.7
  6. #66o3-proOpenAI · $20 in147.4
  7. #67Qwen3.7-PlusAlibaba · $0.32 in147.4

Sources