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.

SpecificationR1DeepSeekClaude Opus 5Anthropic
ReleasedJan 20, 2025Jul 24, 2026
Context window164K1M
Max output16K128K
Input / 1M$0.7$5
Output / 1M$2.5$25
Cost per task$0.24$2.25
Arena Elo—1,393
Serving providers25
Parameters684.5BUndisclosed
LicencemitProprietary
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

  • R1
  • Claude Opus 5

  • Intelligence Index

    R1
    20.4
    Claude Opus 5
    63.1

  • Coding Index

    R1not measured
    Claude Opus 5
    78.0

  • Agentic Index

    R1
    3.1
    Claude Opus 5
    59.2

  • Output speed

    R1
    22 t/s
    Claude Opus 5
    87 t/s

  • Context window

    R1
    164K
    Claude Opus 5
    1M

  • Latency · lower is better

    R1
    956ms
    Claude Opus 5
    1.12s

  • Blended price / 1M · lower is better

    R1
    $1.15
    Claude Opus 5
    $10

  • Cost per task · lower is better

    R1
    $0.24
    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
MetricR1Claude Opus 5
Intelligence Index20.463.1
Coding Index—78.0
Agentic Index3.159.2
Output speed22 t/s87 t/s
Context window164K1M
Latency · lower is better956ms1.12s
Blended price / 1M · lower is better$1.15$10
Cost per task · lower is better$0.24$2.25

Evaluation scores

Percentage correct on a common 0–100% scale

  • R1
  • Claude Opus 5

  • GPQA Diamond

    R1
    81.3%
    Claude Opus 5
    93.2%

  • Humanity's Last Exam

    R1
    15.8%
    Claude Opus 5
    54.9%

  • SciCode

    R1
    40.3%
    Claude Opus 5
    55.7%

  • τ²-bench

    R1
    36.5%
    Claude Opus 5
    42.1%

  • Terminal-Bench Hard

    R1
    15.9%
    Claude Opus 5not measured

  • LiveCodeBench

    R1
    77.0%
    Claude Opus 5not measured

  • AA-LCR long context

    R1
    56.7%
    Claude Opus 5
    75.7%

  • AIME 2025

    R1
    76.0%
    Claude Opus 5not measured

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

View as table
EvaluationR1Claude Opus 5
GPQA Diamond81.3%93.2%
Humanity's Last Exam15.8%54.9%
SciCode40.3%55.7%
τ²-bench36.5%42.1%
Terminal-Bench Hard15.9%—
LiveCodeBench77.0%—
AA-LCR long context56.7%75.7%
AIME 202576.0%—

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