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.

SpecificationM2.5MiniMaxClaude Opus 5Anthropic
ReleasedFeb 12, 2026Jul 24, 2026
Context window205K1M
Max output197K128K
Input / 1M$0.22$5
Output / 1M$0.9$25
Cost per task$0.08$2.25
Arena Elo1,393
Serving providers95
Parameters228.7BUndisclosed
LicenceotherProprietary
Capabilities
  • Reasoning
  • 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

  • M2.5
  • Claude Opus 5
  • Intelligence Index

    M2.5
    34.5
    Claude Opus 5
    63.1
  • Coding Index

    M2.5not measured
    Claude Opus 5
    78.0
  • Agentic Index

    M2.5not measured
    Claude Opus 5
    59.2
  • Output speed

    M2.5
    91 t/s
    Claude Opus 5
    87 t/s
  • Context window

    M2.5
    205K
    Claude Opus 5
    1M
  • Latency · lower is better

    M2.5
    549ms
    Claude Opus 5
    1.12s
  • Blended price / 1M · lower is better

    M2.5
    $0.39
    Claude Opus 5
    $10
  • Cost per task · lower is better

    M2.5
    $0.08
    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
MetricM2.5Claude Opus 5
Intelligence Index34.563.1
Coding Index78.0
Agentic Index59.2
Output speed91 t/s87 t/s
Context window205K1M
Latency · lower is better549ms1.12s
Blended price / 1M · lower is better$0.39$10
Cost per task · lower is better$0.08$2.25

Evaluation scores

Percentage correct on a common 0–100% scale

  • M2.5
  • Claude Opus 5
  • GPQA Diamond

    M2.5
    84.8%
    Claude Opus 5
    93.2%
  • Humanity's Last Exam

    M2.5
    20.5%
    Claude Opus 5
    54.9%
  • SciCode

    M2.5
    42.6%
    Claude Opus 5
    55.7%
  • τ²-bench

    M2.5
    95.3%
    Claude Opus 5
    42.1%
  • Terminal-Bench Hard

    M2.5
    34.8%
    Claude Opus 5not measured
  • AA-LCR long context

    M2.5
    72.0%
    Claude Opus 5
    75.7%

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

View as table
EvaluationM2.5Claude Opus 5
GPQA Diamond84.8%93.2%
Humanity's Last Exam20.5%54.9%
SciCode42.6%55.7%
τ²-bench95.3%42.1%
Terminal-Bench Hard34.8%
LiveCodeBench
AA-LCR long context72.0%75.7%
AIME 2025

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