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.

SpecificationClaude Opus 4AnthropicClaude Opus 5Anthropic
ReleasedMay 22, 2025Jul 24, 2026
Context window200K1M
Max output32K128K
Input / 1M$15$5
Output / 1M$75$25
Cost per task$6.75$2.25
Arena Elo1,2121,393
Serving providers15
ParametersUndisclosedUndisclosed
LicenceProprietaryProprietary
Capabilities
  • Reasoning
  • Tool use
  • Vision
  • 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

  • Claude Opus 4
  • Claude Opus 5

  • Intelligence Index

    Claude Opus 4
    26.0
    Claude Opus 5
    63.1

  • Coding Index

    Claude Opus 4not measured
    Claude Opus 5
    78.0

  • Agentic Index

    Claude Opus 4not measured
    Claude Opus 5
    59.2

  • Output speed

    Claude Opus 4
    22 t/s
    Claude Opus 5
    87 t/s

  • Context window

    Claude Opus 4
    200K
    Claude Opus 5
    1M

  • Latency · lower is better

    Claude Opus 4
    2.22s
    Claude Opus 5
    1.12s

  • Blended price / 1M · lower is better

    Claude Opus 4
    $30
    Claude Opus 5
    $10

  • Cost per task · lower is better

    Claude Opus 4
    $6.75
    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
MetricClaude Opus 4Claude Opus 5
Intelligence Index26.063.1
Coding Index—78.0
Agentic Index—59.2
Output speed22 t/s87 t/s
Context window200K1M
Latency · lower is better2.22s1.12s
Blended price / 1M · lower is better$30$10
Cost per task · lower is better$6.75$2.25

Evaluation scores

Percentage correct on a common 0–100% scale

  • Claude Opus 4
  • Claude Opus 5

  • GPQA Diamond

    Claude Opus 4
    70.1%
    Claude Opus 5
    93.2%

  • Humanity's Last Exam

    Claude Opus 4
    6.2%
    Claude Opus 5
    54.9%

  • SciCode

    Claude Opus 4
    40.9%
    Claude Opus 5
    55.7%

  • τ²-bench

    Claude Opus 4not measured
    Claude Opus 5
    42.1%

  • LiveCodeBench

    Claude Opus 4
    54.2%
    Claude Opus 5not measured

  • AA-LCR long context

    Claude Opus 4
    40.0%
    Claude Opus 5
    75.7%

  • AIME 2025

    Claude Opus 4
    36.3%
    Claude Opus 5not measured

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

View as table
EvaluationClaude Opus 4Claude Opus 5
GPQA Diamond70.1%93.2%
Humanity's Last Exam6.2%54.9%
SciCode40.9%55.7%
τ²-bench—42.1%
Terminal-Bench Hard——
LiveCodeBench54.2%—
AA-LCR long context40.0%75.7%
AIME 202536.3%—

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