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.

SpecificationQwen2.5 Coder 32B InstructQwenClaude Opus 5Anthropic
ReleasedNov 11, 2024Jul 24, 2026
Context window33K1M
Max output33K128K
Input / 1M$0.66$5
Output / 1M$1$25
Cost per task$0.06$2.25
Arena Elo—1,393
Serving providers15
Parameters32.8BUndisclosed
Licenceapache-2.0Proprietary
Capabilities
  • 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

  • Qwen2.5 Coder 32B Instruct
  • Claude Opus 5

  • Intelligence Index

    Qwen2.5 Coder 32B Instruct
    6.9
    Claude Opus 5
    63.1

  • Coding Index

    Qwen2.5 Coder 32B Instructnot measured
    Claude Opus 5
    78.0

  • Agentic Index

    Qwen2.5 Coder 32B Instructnot measured
    Claude Opus 5
    59.2

  • Output speed

    Qwen2.5 Coder 32B Instruct
    26 t/s
    Claude Opus 5
    87 t/s

  • Context window

    Qwen2.5 Coder 32B Instruct
    33K
    Claude Opus 5
    1M

  • Latency · lower is better

    Qwen2.5 Coder 32B Instruct
    484ms
    Claude Opus 5
    1.12s

  • Blended price / 1M · lower is better

    Qwen2.5 Coder 32B Instruct
    $0.745
    Claude Opus 5
    $10

  • Cost per task · lower is better

    Qwen2.5 Coder 32B Instruct
    $0.06
    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
MetricQwen2.5 Coder 32B InstructClaude Opus 5
Intelligence Index6.963.1
Coding Index—78.0
Agentic Index—59.2
Output speed26 t/s87 t/s
Context window33K1M
Latency · lower is better484ms1.12s
Blended price / 1M · lower is better$0.745$10
Cost per task · lower is better$0.06$2.25

Evaluation scores

Percentage correct on a common 0–100% scale

  • Qwen2.5 Coder 32B Instruct
  • Claude Opus 5

  • GPQA Diamond

    Qwen2.5 Coder 32B Instruct
    41.7%
    Claude Opus 5
    93.2%

  • Humanity's Last Exam

    Qwen2.5 Coder 32B Instruct
    3.5%
    Claude Opus 5
    54.9%

  • SciCode

    Qwen2.5 Coder 32B Instruct
    27.1%
    Claude Opus 5
    55.7%

  • τ²-bench

    Qwen2.5 Coder 32B Instructnot measured
    Claude Opus 5
    42.1%

  • LiveCodeBench

    Qwen2.5 Coder 32B Instruct
    29.5%
    Claude Opus 5not measured

  • AA-LCR long context

    Qwen2.5 Coder 32B Instructnot measured
    Claude Opus 5
    75.7%

  • AIME 2025

    Qwen2.5 Coder 32B Instruct
    12.0%
    Claude Opus 5not measured

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

View as table
EvaluationQwen2.5 Coder 32B InstructClaude Opus 5
GPQA Diamond41.7%93.2%
Humanity's Last Exam3.5%54.9%
SciCode27.1%55.7%
τ²-bench—42.1%
Terminal-Bench Hard——
LiveCodeBench29.5%—
AA-LCR long context—75.7%
AIME 202512.0%—

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