RLTasksby ulam.ai
Models / Anthropic

Anthropic / MODEL EVIDENCE

Claude Opus 4

39 benchmarks · 65 observations · 43 sources · released 2025-05-22

Scores at a glance

39 benchmarks · large dot: Claude Opus 4; small dots: closest models where reported

MATH Level 5 ↗85.05%
3.2998.13

Claude Opus 4 · MATH level 5 · Best score (across scorers) · % ↗

Cybench ↗38%
555

Claude Opus 4 · Cybench · Unguided % Solved · % ↗

GSO ↗6.86%
1.344.12

Claude Opus 4 · GSO-Bench · Score OPT@1 · % ↗

1 more reported result

6.9% · Claude Opus 4 · GSO-Bench · Score OPT@1 · % · Epoch AI · GSO-Bench ↗

METR task performance & time horizons ↗63.93%
10.185.2

Claude Opus 4 · METR task performance & time horizons · Average task accuracy ↗

3 more reported results

61.55% · Claude Opus 4 · METR task performance & time horizons · Average task accuracy · Epoch AI · METR Time Horizons ↗

100.36612 minutes · Claude Opus 4 · METR task performance & time horizons · 50% success task horizon · METR-Horizon-v1.1 · Epoch AI · METR Autonomy / Task-Horizon Suite · public result data ↗

85.55663 minutes · Claude Opus 4 · METR task performance & time horizons · 50% success task horizon · Epoch AI · METR Autonomy / Task-Horizon Suite · public result data ↗

LMCA ↗37.43%
2.7868.26

Claude Opus 4 · LMCA · Score · % ↗

DTBench ↗81.6%
41.6598.93

Claude Opus 4 · DTBench · Accuracy · % ↗

LegalBench ↗83.07 %
32.9788.56

Claude Opus 4 (Nonthinking) · LegalBench · Accuracy ↗

MATH-500 ↗90.4 %
25.496.4

Claude Opus 4 (Nonthinking) · MATH 500 · Accuracy ↗

MMLU-Pro ↗86.17 %
30.2892.38

Claude Opus 4 (Nonthinking) · MMLU Pro · Accuracy ↗

MMMU-Pro ↗73.31 %
22.7790.64

Claude Opus 4 (Nonthinking) · MMMU Pro · Accuracy ↗

Fortress ↗27.61%
5.7874.39

Claude Opus 4 · Scale Labs · Risk elicitation score ↗

AlgoTune ↗1.33 ×
1.312.05

Claude Opus 4 (high) · Harmonic-mean speedup ↗

Closest models by score

Similar positions on shared benchmark leaderboards. Smaller gaps are closer.

How similarity is measured

Mean absolute gap in percentile positions on shared benchmark settings; lower is closer. At least 60% overlap is required, with a minimum of three when available. Independent evaluations are used when there are at least three; otherwise published report comparisons are used. Repeated model observations within one setting are averaged before ranking. Writing-style traits and composite indices are excluded. This describes similarity, not an overall quality ranking.

Filter comparisons to this model →

Sources

Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.

  1. Aider LLM Leaderboards — Polyglot coding benchmark ↗

    Aider

  2. ARC Prize - Leaderboard ↗

    ARC Prize Foundation

  3. TextQuests: How Good are LLMs at Text-Based Video Games? — published result table ↗

    Phan, Long, Mazeika, Mantas, Zou, Andy, Hendrycks, Dan · 2025

  4. DTBench — Conceptual Reasoning Index ↗

    Conceptual Reasoning Index

  5. LMCA — Conceptual Reasoning Index ↗

    Conceptual Reasoning Index

  6. Cybench: A Framework for Evaluating Cybersecurity Capabilities and Risks of Language Models ↗

    Zhang, Andy K., Perry, Neil, Dulepet, Riya, Ji, Joey, Menders, Celeste, Lin, Justin W. et al. · 2024

    All 27 authors

    Zhang, Andy K., Perry, Neil, Dulepet, Riya, Ji, Joey, Menders, Celeste, Lin, Justin W., Jones, Eliot, Hussein, Gashon, Liu, Samantha, Jasper, Donovan, Peetathawatchai, Pura, Glenn, Ari, Sivashankar, Vikram, Zamoshchin, Daniel, Glikbarg, Leo, Askaryar, Derek, Yang, Mike, Zhang, Teddy, Alluri, Rishi, Tran, Nathan, Sangpisit, Rinnara, Yiorkadjis, Polycarpos, Osele, Kenny, Raghupathi, Gautham, Boneh, Dan, Ho, Daniel E., Liang, Percy

  7. FutureSearch Evals ↗

    FutureSearch

  8. Algotune — methodology and published results ↗

    Epoch AI · 2026

  9. ARC-AGI-2 ↗

    Epoch AI

  10. CritPt v1 ↗

    Epoch AI

  11. FrontierMath Tier 4 (v1, old) ↗

    Epoch AI

  12. FrontierMath Tiers 1-3 (v1, old) ↗

    Epoch AI

  13. GPQA Diamond ↗

    Epoch AI

  14. Humanity's Last Exam ↗

    Epoch AI

  15. MATH Level 5 ↗

    Epoch AI

  16. METR Time Horizons ↗

    Epoch AI

  17. OTIS Mock AIME 2024-2025 ↗

    Epoch AI

  18. SimpleBench ↗

    Epoch AI

  19. SWE-bench Verified ↗

    Epoch AI

  20. Capabilities & benchmarking — retained external evaluation records ↗

    Epoch AI · 2026

  21. AI benchmarks & capabilities ↗

    Epoch AI

  22. Fiction.liveBench — March 14, 2025 benchmark release ↗

    Fiction.live · 2025

  23. GeoBench — GeoGuessr model benchmark ↗

    GeoBench

  24. Writing benchmark — incorporating mandatory story elements ↗

    lechmazur

  25. GSO: Challenging Software Optimization Tasks for Evaluating SWE-Agents ↗

    Shetty, Manish, Jain, Naman, Liu, Jinjian, Kethanaboyina, Vijay, Sen, Koushik, Stoica, Ion · 2025

  26. WeirdML ↗

    Håvard Tveit Ihle

  27. EnigmaEval ↗

    Scale Labs / SEAL

  28. Scale Labs — evaluation methodology and results ↗

    Scale Labs / SEAL · 2026

  29. Scale Labs — evaluation methodology and results ↗

    Scale Labs / SEAL · 2026

  30. Scale Labs — evaluation methodology and results ↗

    Scale Labs / SEAL · 2026

  31. Measuring AI Ability to Complete Long Software Tasks ↗

    METR

  32. Metr Autonomy Task Horizon Suite — methodology and published results ↗

    METR · 2026

  33. AIME — evaluation methodology and results ↗

    Vals AI · 2026

  34. GPQA Diamond — evaluation methodology and results ↗

    Vals AI · 2026

  35. LiveCodeBench — evaluation methodology and results ↗

    Vals AI · 2026

  36. LegalBench — evaluation methodology and results ↗

    Vals AI · 2026

  37. MATH 500 — evaluation methodology and results ↗

    Vals AI · 2026

  38. MedQA — evaluation methodology and results ↗

    Vals AI · 2026

  39. MGSM — evaluation methodology and results ↗

    Vals AI · 2026

  40. MMLU Pro — evaluation methodology and results ↗

    Vals AI · 2026

  41. MMMU Pro — evaluation methodology and results ↗

    Vals AI · 2026

  42. MortgageTax — evaluation methodology and results ↗

    Vals AI · 2026

  43. TaxEval — evaluation methodology and results ↗

    Vals AI · 2026

Names used in source tables

Claude Opus 4 · Claude Opus 4 · Claude Opus 4 (Nonthinking) · Claude Opus 4 (Thinking) · Claude Opus 4 (high)