Skip to content
llmwaves

Compare

Four models, one page

Pick the models you are actually choosing between. Every row is scaled within itself, so a bar means something next to its neighbours and nothing across rows.

SpecificationKimi K2 0905Moonshot AIClaude Opus 5Anthropic
ReleasedSep 4, 2025Jul 24, 2026
Context window262K1M
Max output100K128K
Input / 1M$0.6$5
Output / 1M$2.5$25
Cost per task$0.09$2.25
Arena Elo—1,393
Serving providers15
Parameters1026.5BUndisclosed
LicenceotherProprietary
Capabilities
  • Tool use
  • Open weights
  • Reasoning
  • Tool use
  • Vision

Metrics side by side

Each row is scaled to the largest value in that row — bars compare within a row, never across rows

  • Kimi K2 0905
  • Claude Opus 5

  • Intelligence Index

    Kimi K2 0905
    24.0
    Claude Opus 5
    63.1

  • Coding Index

    Kimi K2 0905not measured
    Claude Opus 5
    78.0

  • Agentic Index

    Kimi K2 0905not measured
    Claude Opus 5
    59.2

  • Output speed

    Kimi K2 0905
    19 t/s
    Claude Opus 5
    87 t/s

  • Context window

    Kimi K2 0905
    262K
    Claude Opus 5
    1M

  • Latency · lower is better

    Kimi K2 0905
    861ms
    Claude Opus 5
    1.12s

  • Blended price / 1M · lower is better

    Kimi K2 0905
    $1.07
    Claude Opus 5
    $10

  • Cost per task · lower is better

    Kimi K2 0905
    $0.09
    Claude Opus 5
    $2.25

Rows marked “lower is better” still draw a longer bar for a larger number — read the value, not just the length. Arena Elo is in the specification table above instead: it has no meaningful zero, so a bar would flatten the gaps.

View as table
MetricKimi K2 0905Claude Opus 5
Intelligence Index24.063.1
Coding Index—78.0
Agentic Index—59.2
Output speed19 t/s87 t/s
Context window262K1M
Latency · lower is better861ms1.12s
Blended price / 1M · lower is better$1.07$10
Cost per task · lower is better$0.09$2.25

Evaluation scores

Percentage correct on a common 0–100% scale

  • Kimi K2 0905
  • Claude Opus 5

  • GPQA Diamond

    Kimi K2 0905
    76.7%
    Claude Opus 5
    93.2%

  • Humanity's Last Exam

    Kimi K2 0905
    6.4%
    Claude Opus 5
    54.9%

  • SciCode

    Kimi K2 0905
    30.7%
    Claude Opus 5
    55.7%

  • τ²-bench

    Kimi K2 0905
    73.4%
    Claude Opus 5
    42.1%

  • Terminal-Bench Hard

    Kimi K2 0905
    23.5%
    Claude Opus 5not measured

  • LiveCodeBench

    Kimi K2 0905
    61.0%
    Claude Opus 5not measured

  • AA-LCR long context

    Kimi K2 0905
    53.7%
    Claude Opus 5
    75.7%

  • AIME 2025

    Kimi K2 0905
    57.3%
    Claude Opus 5not measured

A missing bar means that evaluation was not run for that model — it is not a zero.

View as table
EvaluationKimi K2 0905Claude Opus 5
GPQA Diamond76.7%93.2%
Humanity's Last Exam6.4%54.9%
SciCode30.7%55.7%
τ²-bench73.4%42.1%
Terminal-Bench Hard23.5%—
LiveCodeBench61.0%—
AA-LCR long context53.7%75.7%
AIME 202557.3%—

Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).