RLTasksby ulam.ai
Models / Anthropic

Anthropic / MODEL EVIDENCE

Claude Opus 5

93 benchmarks · 167 observations · 83 sources · released 2026-07-24

Scores at a glance

93 benchmarks · large dot: Claude Opus 5; small dots: closest models where reported

GPQA Diamond ↗92.93%
92.9392.93

Claude Opus 5 · GPQA diamond · Best score (across scorers) · % ↗

5 more reported results

93.43 % · Claude Opus 5 · GPQA Diamond · Accuracy · Vals AI · GPQA Diamond ↗

93.74% · Claude Opus 5 (Xhigh) · GPQA Diamond · AA implementation · Artificial Analysis ↗

93.88% · Claude Opus 5 (max) · GPQA diamond · Best score (across scorers) · % · Epoch AI · GPQA diamond ↗

87.88% · Claude Opus 5 (low) · GPQA diamond · Best score (across scorers) · % · Epoch AI · GPQA diamond ↗

93.7% · Claude Opus 5 · GPQA Diamond · GPT-6 Astra launch report ↗

DeepSWE v1.1 ↗73.7%
67.474.1

Claude Opus 5 · DeepSWE v1.1 ↗

6 more reported results

74.0 · Opus 5 · DeepSWE v1.1 · DeepSeek V4.1 Flash 1017 model card ↗

73.65% · Claude Opus 5 (max) · DeepSWE · Pass@1 · % · Epoch AI · DeepSWE ↗

73.15% · Claude Opus 5 (xhigh) · DeepSWE · Pass@1 · % · Epoch AI · DeepSWE ↗

72.83% · Claude Opus 5 (high) · DeepSWE · Pass@1 · % · Epoch AI · DeepSWE ↗

68.9% · Claude Opus 5 (medium) · DeepSWE · Pass@1 · % · Epoch AI · DeepSWE ↗

58.13% · Claude Opus 5 (low) · DeepSWE · Pass@1 · % · Epoch AI · DeepSWE ↗

CursorBench 4.0 ↗46.6%
41.757.8

Opus 5 · CursorBench 4.0 ↗

5 more reported results

46.6% · Claude Opus 5 (max) · CursorBench · Score · % · Epoch AI · CursorBench ↗

46.1% · Claude Opus 5 (xhigh) · CursorBench · Score · % · Epoch AI · CursorBench ↗

44.7% · Claude Opus 5 (high) · CursorBench · Score · % · Epoch AI · CursorBench ↗

43.3% · Claude Opus 5 (medium) · CursorBench · Score · % · Epoch AI · CursorBench ↗

40.7% · Claude Opus 5 (low) · CursorBench · Score · % · Epoch AI · CursorBench ↗

OSWorld 2.0 ↗70.2%†
65.772.6

Claude Opus 5 · OSWorld 2.0 · offline, partial ↗

5 more reported results

31.43% · Claude Opus 5 (max) · OSWorld 2.0 · Binary accuracy · % · Epoch AI · OSWorld 2.0 ↗

30.21% · Claude Opus 5 (xhigh) · OSWorld 2.0 · Binary accuracy · % · Epoch AI · OSWorld 2.0 ↗

28.98% · Claude Opus 5 (high) · OSWorld 2.0 · Binary accuracy · % · Epoch AI · OSWorld 2.0 ↗

25.33% · Claude Opus 5 (medium) · OSWorld 2.0 · Binary accuracy · % · Epoch AI · OSWorld 2.0 ↗

22.32% · Claude Opus 5 (low) · OSWorld 2.0 · Binary accuracy · % · Epoch AI · OSWorld 2.0 ↗

HealthBench ↗56.4%†
52.163.4

Claude Opus 5 · HealthBench · HealthBench Professional · length-adjusted ↗

BALROG ↗63.4%
3.768.3

Claude Opus 5 (max) · Balrog · Average progress · % ↗

GDP.pdf ↗24%
235

Claude Opus 5 (max) · GDP.pdf · GDP.pdf score · % ↗

Vals RSI Index ↗0.43 reference score
00.46

Claude Opus 5 (max) · Compression · reference-anchored score ↗

3 more reported results

0.43 reference score · Claude Opus 5 (max) · LM training · reference-anchored score · Vals RSI Index · LM training ↗

0.15 reference score · Claude Opus 5 (max) · Harness engineering · reference-anchored score · Vals RSI Index · Harness engineering ↗

0.32 reference score · Claude Opus 5 (max) · Post-training · reference-anchored score · Vals RSI Index · Post-training ↗

Fortress ↗16.84%
5.7874.39

claude-opus-5 · Scale Labs · Risk elicitation score ↗

Closest models by score

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

17.6 percentile-point gap

41 of 52 shared benchmark comparisons

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 →

Reports featuring this model

DeepSeek V4.1 Flash 1017 · DeepSeek →

Claude Opus 5.5 · Anthropic →

GPT-6 Astra · OpenAI →

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. ARC Prize - Leaderboard ↗

    ARC Prize Foundation

  2. Arena leaderboard — WebDev evaluation ↗

    Arena

  3. AA-AnalystAgent Benchmark | Artificial Analysis ↗

    Artificial Analysis

  4. EnterpriseOps-Gym-AA Benchmark | Artificial Analysis ↗

    Artificial Analysis

  5. GPQA Diamond Benchmark | Artificial Analysis ↗

    Artificial Analysis

  6. MMMU-Pro Benchmark | Artificial Analysis ↗

    Artificial Analysis

  7. SciCode Benchmark | Artificial Analysis ↗

    Artificial Analysis

  8. 𝜏³-Banking Benchmark | Artificial Analysis ↗

    Artificial Analysis

  9. Terminal-Bench 2.1 Benchmark | Artificial Analysis ↗

    Artificial Analysis

  10. BALROG: Benchmarking Agentic LLM and VLM Reasoning On Games ↗

    Paglieri, Davide, Cupiał, Bartłomiej, Coward, Samuel, Piterbarg, Ulyana, Wolczyk, Maciej, Khan, Akbir et al. · 2024

    All 13 authors

    Paglieri, Davide, Cupiał, Bartłomiej, Coward, Samuel, Piterbarg, Ulyana, Wolczyk, Maciej, Khan, Akbir, Pignatelli, Eduardo, Kuciński, Łukasz, Pinto, Lerrel, Fergus, Rob, Foerster, Jakob Nicolaus, Parker-Holder, Jack, Rocktäschel, Tim

  11. FrontierCode Leaderboard ↗

    Cognition

  12. DTBench — Conceptual Reasoning Index ↗

    Conceptual Reasoning Index

  13. LMCA — Conceptual Reasoning Index ↗

    Conceptual Reasoning Index

  14. Cursor · CursorBench ↗

    Cursor

  15. DeepSWE ↗

    Datacurve

  16. APEX-Agents ↗

    Epoch AI

  17. ARC-AGI-2 ↗

    Epoch AI

  18. Chess Puzzles ↗

    Epoch AI

  19. CritPt v1 ↗

    Epoch AI

  20. Earthborne Rangers (EBR-bench) ↗

    Epoch AI

  21. FrontierMath Erdős ↗

    Epoch AI

  22. FrontierMath Tier 4 (v2) ↗

    Epoch AI

  23. FrontierMath Tiers 1-3 (v2) ↗

    Epoch AI

  24. Furniture Assembly Benchmark ↗

    Epoch AI

  25. GPQA Diamond ↗

    Epoch AI

  26. Mystery Game Puzzles ↗

    Epoch AI

  27. OTIS Mock AIME 2024-2025 ↗

    Epoch AI

  28. SimpleQA Verified ↗

    Epoch AI

  29. Vending Bench 2 — methodology and published results ↗

    Epoch AI · 2026

  30. AI benchmarks & capabilities ↗

    Epoch AI

  31. WeirdML ↗

    Håvard Tveit Ihle

  32. deepseek-ai/DeepSeek-V4.1-Flash-1017 · Hugging Face ↗

    DeepSeek AI

  33. DrugDiscoveryBench — evaluation methodology and results ↗

    Scale Labs / SEAL · 2026

  34. Scale Labs — evaluation methodology and results ↗

    Scale Labs / SEAL · 2026

  35. HiL-Bench (Human-in-Loop Benchmark) — evaluation methodology and results ↗

    Scale Labs / SEAL · 2026

  36. Scale Labs — evaluation methodology and results ↗

    Scale Labs / SEAL · 2026

  37. Humanity's Sixth Sense — evaluation methodology and results ↗

    Scale Labs / SEAL · 2026

  38. Scale Labs — evaluation methodology and results ↗

    Scale Labs / SEAL · 2026

  39. SWE-Bench Pro V2 — evaluation methodology and results ↗

    Scale Labs / SEAL · 2026

  40. SWE-Bench Pro V2 — evaluation methodology and results ↗

    Scale Labs / SEAL · 2026

  41. Scale Labs — evaluation methodology and results ↗

    Scale Labs / SEAL · 2026

  42. SWE Atlas - Test Writing — evaluation methodology and results ↗

    Scale Labs / SEAL · 2026

  43. GPT-6 Astra: A new generation of intelligence ↗

    OpenAI

  44. OSWorld 2.0: Benchmarking computer-use agents on long-horizon real-world tasks ↗

    OSWorld Team

  45. GDP.pdf Benchmark | Surge AI ↗

    Surge AI

  46. Introducing Claude Opus 5.5 ↗

    Anthropic

  47. FrontierSWE — software engineering benchmark ↗

    FrontierSWE

  48. BioMysteryBench — evaluation methodology and results ↗

    Vals AI · 2026

  49. Code Migration v1 — evaluation methodology and results ↗

    Vals AI · 2026

  50. CorpFin — evaluation methodology and results ↗

    Vals AI · 2026

  51. CUA-bench — evaluation methodology and results ↗

    Vals AI · 2026

  52. CyberBench — evaluation methodology and results ↗

    Vals AI · 2026

  53. Excel Modeling Benchmark v1 — evaluation methodology and results ↗

    Vals AI · 2026

  54. Finance Agent v2 — evaluation methodology and results ↗

    Vals AI · 2026

  55. GPQA Diamond — evaluation methodology and results ↗

    Vals AI · 2026

  56. Harvey's Legal Agent Benchmark v1 — evaluation methodology and results ↗

    Vals AI · 2026

  57. IOI — evaluation methodology and results ↗

    Vals AI · 2026

  58. LiveCodeBench — evaluation methodology and results ↗

    Vals AI · 2026

  59. LegalBench — evaluation methodology and results ↗

    Vals AI · 2026

  60. Legal Research Bench — evaluation methodology and results ↗

    Vals AI · 2026

  61. MedCode — evaluation methodology and results ↗

    Vals AI · 2026

  62. MedScribe — evaluation methodology and results ↗

    Vals AI · 2026

  63. MMLU Pro — evaluation methodology and results ↗

    Vals AI · 2026

  64. MMMU Pro — evaluation methodology and results ↗

    Vals AI · 2026

  65. MortgageTax — evaluation methodology and results ↗

    Vals AI · 2026

  66. MysteryMechanism — evaluation methodology and results ↗

    Vals AI · 2026

  67. ProgramBench — evaluation methodology and results ↗

    Vals AI · 2026

  68. ProofBench — evaluation methodology and results ↗

    Vals AI · 2026

  69. Public Benefits Bench — evaluation methodology and results ↗

    Vals AI · 2026

  70. Vals RSI Index — evaluation methodology and results ↗

    Vals AI · 2026

  71. SAGE — evaluation methodology and results ↗

    Vals AI · 2026

  72. Vals Science Index — evaluation methodology and results ↗

    Vals AI · 2026

  73. SkillsBench — evaluation methodology and results ↗

    Vals AI · 2026

  74. SRE Bench — evaluation methodology and results ↗

    Vals AI · 2026

  75. SWE-bench Verified — evaluation methodology and results ↗

    Vals AI · 2026

  76. Tax Agent Bench — evaluation methodology and results ↗

    Vals AI · 2026

  77. TaxEval — evaluation methodology and results ↗

    Vals AI · 2026

  78. Terminal-Bench 4.0 — evaluation methodology and results ↗

    Vals AI · 2026

  79. Terminal-Bench-Science — evaluation methodology and results ↗

    Vals AI · 2026

  80. Time Horizon Index: KSP — evaluation methodology and results ↗

    Vals AI · 2026

  81. Vals Multimodal Index — evaluation methodology and results ↗

    Vals AI · 2026

  82. Vibe Code Bench 1-100 — evaluation methodology and results ↗

    Vals AI · 2026

  83. Vibe Code Bench — evaluation methodology and results ↗

    Vals AI · 2026

Names used in source tables

Claude Opus 5 · Claude Opus 5 (Claude Code) max · Claude Opus 5 (High) · Claude Opus 5 (Max) · Claude Opus 5 (Mini-SWE-agent) · Claude Opus 5 (OpenHands) · Claude Opus 5 (Xhigh) · Claude Opus 5 (high) · Claude Opus 5 (low) · Claude Opus 5 (max) · Claude Opus 5 (medium) · Claude Opus 5 (xhigh) · Claude-Opus-5 (max) · Opus 5 · Opus 5 (Claude Code) xHigh · Opus 5 (Claude Code) xhigh · claude-opus-5 · claude-opus-5 (xhigh)