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.

SpecificationGranite 4.1 8BIBM GraniteClaude Opus 5Anthropic
ReleasedApr 30, 2026Jul 24, 2026
Context window131K1M
Max output131K128K
Input / 1M$0.05$5
Output / 1M$0.1$25
Cost per task$0.005$2.25
Arena Elo—1,393
Serving providers15
Parameters8.8BUndisclosed
Licenceapache-2.0Proprietary
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

  • Granite 4.1 8B
  • Claude Opus 5

  • Intelligence Index

    Granite 4.1 8B
    6.4
    Claude Opus 5
    63.1

  • Coding Index

    Granite 4.1 8B
    9.5
    Claude Opus 5
    78.0

  • Agentic Index

    Granite 4.1 8Bnot measured
    Claude Opus 5
    59.2

  • Output speed

    Granite 4.1 8B
    121 t/s
    Claude Opus 5
    87 t/s

  • Context window

    Granite 4.1 8B
    131K
    Claude Opus 5
    1M

  • Latency · lower is better

    Granite 4.1 8B
    118ms
    Claude Opus 5
    1.12s

  • Blended price / 1M · lower is better

    Granite 4.1 8B
    $0.063
    Claude Opus 5
    $10

  • Cost per task · lower is better

    Granite 4.1 8B
    $0.005
    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
MetricGranite 4.1 8BClaude Opus 5
Intelligence Index6.463.1
Coding Index9.578.0
Agentic Index—59.2
Output speed121 t/s87 t/s
Context window131K1M
Latency · lower is better118ms1.12s
Blended price / 1M · lower is better$0.063$10
Cost per task · lower is better$0.005$2.25

Evaluation scores

Percentage correct on a common 0–100% scale

  • Granite 4.1 8B
  • Claude Opus 5

  • GPQA Diamond

    Granite 4.1 8B
    43.3%
    Claude Opus 5
    93.2%

  • Humanity's Last Exam

    Granite 4.1 8B
    3.8%
    Claude Opus 5
    54.9%

  • SciCode

    Granite 4.1 8B
    21.8%
    Claude Opus 5
    55.7%

  • τ²-bench

    Granite 4.1 8B
    27.8%
    Claude Opus 5
    42.1%

  • Terminal-Bench Hard

    Granite 4.1 8B
    0.0%
    Claude Opus 5not measured

  • AA-LCR long context

    Granite 4.1 8B
    11.3%
    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
EvaluationGranite 4.1 8BClaude Opus 5
GPQA Diamond43.3%93.2%
Humanity's Last Exam3.8%54.9%
SciCode21.8%55.7%
τ²-bench27.8%42.1%
Terminal-Bench Hard0.0%—
LiveCodeBench——
AA-LCR long context11.3%75.7%
AIME 2025——

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