Opus 5: score and benchmark evidence

Historical release · · Anthropic

125.7 points · Rank 3 among 19 ranked models · Stability range 119.4–131.7 points.

Rank follows the point estimate; it does not establish statistically significant superiority. The scale is anchored at mean 100, SD 15 in the frozen calibration cohort, not human IQ or a percentage.

One capability score combines evidence across seven equally weighted domains. Related benchmark variants share a family budget, and influence from one evaluator is capped.

Release tli-2026-v5.1-2026-09-06. Methodology and limitations for this release.

This page preserves historical evidence. See the current model record.

Capability domains

Domain scores use calibration-cohort standard-deviation units. Unmeasured domains are shown as unavailable, not zero.

DomainScore (SD units)Observed benchmarks
Knowledge & reasoning0.817
Coding & software engineering0.847
Agentic tool & computer use0.564
Professional & real-world work0.8311
Multimodal & vision0.683
CybersecurityUnavailable0
Preference & communication0.551

Published benchmark evidence

33 results contribute to this score, from 94 collected results and 31 contributing benchmark families. Counts are not independent sample sizes.

Only results with a matched source record, metric and documented configuration contribute. Developer-reported results are eligible when their configuration is clear. Unresolved records remain in the database with their exclusion reason. Archived reviews and live source checks are labelled separately.

The bar estimates score stability under the scoring assumptions: a fixed calibration and comparable benchmark availability. It resamples benchmark families and evaluators while keeping each domain’s evidence strength fixed, then adds a frozen correction learned from missing-evidence tests. Benchmark count and domain coverage both affect that correction. It does not add a second full capability-posterior draw or uncertainty from changing the reference calibration. It is not a guarantee under selective publication, new benchmark domains or future model releases.

Reporting sensitivity: ±1.5 points. Sensitivity to selectively published results. This stress range is not a confidence interval or a probability statement.

  1. GDPval-AA v2 (Elo): 1861 native-points

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  2. Terminal-Bench 2.1: 86.7%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  3. DeepSWE v1.1: 74%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  4. LMArena (Chatbot Arena) — Text: 1489 native-points

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  5. AA Intelligence Index v4.1.1 (score): 63 native-points

    Not used in the score. Source category not recorded. Review status: excluded.

    Held-out composite, never a scoring input

  6. GPQA Diamond: 93.2%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: Claude Opus 5 (Adaptive Reasoning, Max Effort); AA published evaluation harness

    Metric: accuracy, AA

    Original source · Retrieved 2026-09-06

    Current first-party evaluator data matches the recorded score at its published rounding. Exact pinned configuration selected; later value changes are not silently substituted.

  7. AA-LCR: 75.7%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  8. Finance Agent v2 — Vals: 58.6%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: anthropic_claude-opus-5; Vals max effort

    Metric: Finance Agent (v2) overall percentage; Vals benchmark-specific harness

    Original source · Retrieved 2026-09-06

    Exact model and benchmark row; current source value agrees at sheet rounding. Uses published default parameters; source warns individual benchmarks may override them. Known mismatched rows/harnesses are not imported.

  9. Excel Modeling Benchmark — Vals: 73.6%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: anthropic_claude-opus-5; Vals max effort

    Metric: EMB overall percentage; Vals benchmark-specific harness

    Original source · Retrieved 2026-09-06

    Exact model and benchmark row; current source value agrees at sheet rounding. Uses published default parameters; source warns individual benchmarks may override them. Known mismatched rows/harnesses are not imported.

  10. SciCode: 56%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  11. Humanity's Last Exam (no tools): 52.6%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  12. Legal Research Bench — Vals: 55.3%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: anthropic_claude-opus-5; Vals max effort

    Metric: Legal Research Bench overall percentage; Vals benchmark-specific harness

    Original source · Retrieved 2026-09-06

    Exact model and benchmark row; current source value agrees at sheet rounding. Uses published default parameters; source warns individual benchmarks may override them. Known mismatched rows/harnesses are not imported.

  13. Harvey LAB (Vals): 6.7%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: anthropic_claude-opus-5; Vals max effort

    Metric: Harvey's Legal Agent Benchmark overall percentage; Vals benchmark-specific harness

    Original source · Retrieved 2026-09-06

    Exact model and benchmark row; current source value agrees at sheet rounding. Uses published default parameters; source warns individual benchmarks may override them. Known mismatched rows/harnesses are not imported.

  14. AA-Briefcase Overall Elo: 1720 native-points

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  15. MedScribe — Vals: 91%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: anthropic_claude-opus-5; Vals max effort

    Metric: MedScribe overall percentage; Vals benchmark-specific harness

    Original source · Retrieved 2026-09-06

    Exact model and benchmark row; current source value agrees at sheet rounding. Uses published default parameters; source warns individual benchmarks may override them. Known mismatched rows/harnesses are not imported.

  16. SWE-bench Verified — Vals: 97%

    Contributes to the score. Source category: independent. Review status: reviewed-archive.

    Configuration: claude-opus-5; max effort; Vals AI mini-SWE-agent bash harness

    Metric: Resolved issue rate

    Original source · Retrieved 2026-08-16T00:00:00.000Z

    Exact-value match to an existing verified, comparable, pinned-configuration observation. Archived review retained; not asserted newly independently replicated.

  17. LegalBench — Vals: 87%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: anthropic_claude-opus-5; Vals max effort

    Metric: LegalBench overall percentage; Vals benchmark-specific harness

    Original source · Retrieved 2026-09-06

    Exact model and benchmark row; current source value agrees at sheet rounding. Uses published default parameters; source warns individual benchmarks may override them. Known mismatched rows/harnesses are not imported.

  18. TaxEval v2 — Vals: 75.1%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: anthropic_claude-opus-5; Vals max effort

    Metric: TaxEval v2 overall percentage; Vals benchmark-specific harness

    Original source · Retrieved 2026-09-06

    Exact model and benchmark row; current source value agrees at sheet rounding. Uses published default parameters; source warns individual benchmarks may override them. Known mismatched rows/harnesses are not imported.

  19. KingBench 3: 77.5%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  20. Vals Index: 74.8%

    Not used in the score. Source category: independent. Review status: excluded.

    Configuration: claude-opus-5; max effort; Vals AI index methodology

    Metric: Composite score

    Original source · Retrieved 2026-08-16T00:00:00.000Z

    Held-out composite, never a scoring input

  21. Public Benefits Bench — Vals: 76.9%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: anthropic_claude-opus-5; Vals max effort

    Metric: Public Benefits Bench v1.1 overall percentage; Vals benchmark-specific harness

    Original source · Retrieved 2026-09-06

    Exact model and benchmark row; current source value agrees at sheet rounding. Uses published default parameters; source warns individual benchmarks may override them. Known mismatched rows/harnesses are not imported.

  22. MedCode — Vals: 63.6%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: anthropic_claude-opus-5; Vals max effort

    Metric: MedCode overall percentage; Vals benchmark-specific harness

    Original source · Retrieved 2026-09-06

    Exact model and benchmark row; current source value agrees at sheet rounding. Uses published default parameters; source warns individual benchmarks may override them. Known mismatched rows/harnesses are not imported.

  23. AutomationBench v1.0.6: 50.3%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  24. AA-Omniscience: 31.27 native-points

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  25. MMLU Pro — Vals: 91.6%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: anthropic_claude-opus-5; Vals max effort

    Metric: MMLU Pro overall percentage; Vals benchmark-specific harness

    Original source · Retrieved 2026-09-06

    Exact model and benchmark row; current source value agrees at sheet rounding. Uses published default parameters; source warns individual benchmarks may override them. Known mismatched rows/harnesses are not imported.

  26. LiveBench: 80.1%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  27. CorpFin v2 — Vals: 73.2%

    Contributes to the score. Source category: independent. Review status: reviewed-archive.

    Configuration: claude-opus-5; max effort; Vals AI CorpFin v2 evaluation with Sonnet 4.5 judge

    Metric: Accuracy

    Original source · Retrieved 2026-08-16T00:00:00.000Z

    Exact-value match to an existing verified, comparable, pinned-configuration observation. Archived review retained; not asserted newly independently replicated.

  28. MortgageTax — Vals: 72.1%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: anthropic_claude-opus-5; Vals max effort

    Metric: MortgageTax overall percentage; Vals benchmark-specific harness

    Original source · Retrieved 2026-09-06

    Exact model and benchmark row; current source value agrees at sheet rounding. Uses published default parameters; source warns individual benchmarks may override them. Known mismatched rows/harnesses are not imported.

  29. SAGE — Vals: 49.4%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: anthropic_claude-opus-5; Vals max effort

    Metric: SAGE overall percentage; Vals benchmark-specific harness

    Original source · Retrieved 2026-09-06

    Exact model and benchmark row; current source value agrees at sheet rounding. Uses published default parameters; source warns individual benchmarks may override them. Known mismatched rows/harnesses are not imported.

  30. Code Migration †: 57.5%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: anthropic_claude-opus-5; Vals max effort

    Metric: Code Migration overall percentage; Vals benchmark-specific harness

    Original source · Retrieved 2026-09-06

    Exact model and benchmark row; current source value agrees at sheet rounding. Uses published default parameters; source warns individual benchmarks may override them. Known mismatched rows/harnesses are not imported.

  31. Toolathlon-Verified: 80.6%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  32. Agents' Last Exam: 31.6%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  33. Terminal-Bench v3.0: 42.7%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  34. LiveCodeBench: 89.03%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: anthropic_claude-opus-5; Vals max effort

    Metric: LiveCodeBench overall percentage; Vals benchmark-specific harness

    Original source · Retrieved 2026-09-06

    Exact model and benchmark row; current source value agrees at sheet rounding. Uses published default parameters; source warns individual benchmarks may override them. Known mismatched rows/harnesses are not imported.

  35. Humanity's Last Exam w/ tools: 64.7%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  36. Design Arena (Elo): 1407 native-points

    Contributes to the score. Source category: independent. Review status: reviewed-archive.

    Configuration: claude-opus-5; max effort; Design Arena 2026-08-11 rating snapshot

    Metric: Elo rating

    Original source · Retrieved 2026-08-16T00:00:00.000Z

    Exact-value match to an existing verified, comparable, pinned-configuration observation. Archived review retained; not asserted newly independently replicated.

  37. Vals Multimodal Index: 73.9%

    Not used in the score. Source category: independent. Review status: excluded.

    Configuration: claude-opus-5; max effort; Vals AI multimodal index methodology

    Metric: Composite score

    Original source · Retrieved 2026-08-16T00:00:00.000Z

    Held-out composite, never a scoring input

  38. SWE-Marathon v1.1: 50%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  39. ProgramBench v1 (Raw Pass Rate) — Vals: 82.273%

    Contributes to the score. Source category: independent. Review status: reviewed-archive.

    Configuration: anthropic/claude-opus-5 · compute=max

    Metric: Raw Pass Rate: mean percent of hidden behavioral tests passed per task. Not Fully Resolved (all tests pass) and not Almost Resolved (≥95% of tests pass).

    Original source · Retrieved 2026-09-05

    Matched exact metric/value and the archived benchmark-specific configuration review. Unknown native effort stays explicitly unknown.

  40. SkillsBench †: 60.4%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  41. FrontierMath v2 Tier 4: 73.17%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  42. CursorBench v3.2: 70%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  43. APEX-Agents Mean Criteria Passed: 60.6%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  44. SWE-bench Pro: 79.2%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  45. CharXiv (with CI / RQ): 89.3%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  46. APEX-SWE (Pass@1, Terminus-2): 54.67%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  47. BrowseComp: 90.8%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  48. MMMU-Pro: 84.7%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: Claude Opus 5 (Adaptive Reasoning, Max Effort); AA published evaluation harness

    Metric: accuracy, AA

    Original source · Retrieved 2026-09-06

    Current first-party evaluator data matches the recorded score at its published rounding. Exact pinned configuration selected; later value changes are not silently substituted.

  49. ARC-AGI-2: 90.4%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  50. OSWorld 2.0 — Anthropic H2H: 70.6%

    Contributes to the score. Source category: third_party. Review status: reviewed-archive.

    Configuration: claude-opus-5; max effort; Anthropic Opus 5 published head-to-head setup

    Metric: Task success

    Original source · Retrieved 2026-08-16T00:00:00.000Z

    Exact-value match to an existing verified, comparable, pinned-configuration observation. Archived review retained; not asserted newly independently replicated.

  51. ExploitBench: 70%

    Not used in the score. Source category not recorded. Review status: excluded.

    Inherited comparator metric/scaffold equivalence to Astra Cap Percent is not established; the source flags possible contamination. Excluded pending protocol audit.

  52. PostTrainBench: 35.04%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  53. SimpleQA: 56.7%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  54. FrontierCode v1.1 Extended: 63.6%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  55. JobBench: 65.7%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  56. OSWorld-Verified: 90.69%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  57. Creative Writing v3: 2116.1 native-points

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  58. MathArena — ArXivMath Jun 2026: 80.95%

    Contributes to the score. Source category: benchmark_org. Review status: reviewed-archive.

    Configuration: claude-opus-5; max effort; MathArena June 2026 competition harness

    Metric: Accuracy

    Original source · Retrieved 2026-08-16T00:00:00.000Z

    Exact-value match to an existing verified, comparable, pinned-configuration observation. Archived review retained; not asserted newly independently replicated.

  59. MathArena — BrokenArXiv Jun 2026: 90.74%

    Contributes to the score. Source category: benchmark_org. Review status: reviewed-archive.

    Configuration: claude-opus-5; max effort; MathArena June 2026 competition harness

    Metric: Accuracy

    Original source · Retrieved 2026-08-16T00:00:00.000Z

    Exact-value match to an existing verified, comparable, pinned-configuration observation. Archived review retained; not asserted newly independently replicated.

  60. AIIQ Composite IQ: 134 native-points

    Not used in the score. Source category: independent. Review status: excluded.

    Configuration: claude-opus-5; max effort; AIIQ composite snapshot

    Metric: Composite IQ

    Original source · Retrieved 2026-08-16T00:00:00.000Z

    Held-out composite, never a scoring input

  61. MCP Atlas †: 85.8%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  62. OfficeQA Pro †: 66.9%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  63. EQ-Bench 4: 1385 native-points

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  64. MathArena — ArXivLean Jun 2026: 31.25%

    Contributes to the score. Source category: benchmark_org. Review status: reviewed-archive.

    Configuration: claude-opus-5; max effort; MathArena June 2026 competition harness

    Metric: Accuracy

    Original source · Retrieved 2026-08-16T00:00:00.000Z

    Exact-value match to an existing verified, comparable, pinned-configuration observation. Archived review retained; not asserted newly independently replicated.

  65. Frontier-Bench v0.1 — Anthropic H2H: 43.3%

    Contributes to the score. Source category: third_party. Review status: reviewed-archive.

    Configuration: claude-opus-5; max effort; Anthropic Opus 5 published head-to-head setup

    Metric: Task success

    Original source · Retrieved 2026-08-16T00:00:00.000Z

    Exact-value match to an existing verified, comparable, pinned-configuration observation. Archived review retained; not asserted newly independently replicated.

  66. FrontierCode v1.1 Main — Anthropic H2H: 53.4%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  67. AutomationBench — Anthropic H2H: 26%

    Contributes to the score. Source category: third_party. Review status: reviewed-archive.

    Configuration: claude-opus-5; max effort; Anthropic Opus 5 published head-to-head setup

    Metric: Task success

    Original source · Retrieved 2026-08-16T00:00:00.000Z

    Exact-value match to an existing verified, comparable, pinned-configuration observation. Archived review retained; not asserted newly independently replicated.

  68. Harvey Legal Agent Benchmark — held-out: 11.7%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  69. τ³-Banking: 42.1%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: Claude Opus 5 (Adaptive Reasoning, Max Effort); AA published evaluation harness

    Metric: strict task success, AA

    Original source · Retrieved 2026-09-06

    Current first-party evaluator data matches the recorded score at its published rounding. Exact pinned configuration selected; later value changes are not silently substituted.

  70. CritPt: 29.1%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: Claude Opus 5 (Adaptive Reasoning, Max Effort); AA published evaluation harness

    Metric: accuracy, AA official grading

    Original source · Retrieved 2026-09-06

    Current first-party evaluator data matches the recorded score at its published rounding. Exact pinned configuration selected; later value changes are not silently substituted.

  71. Terminal-Bench 4.0: 51.8%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  72. GDP.pdf: 24%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  73. ProofBench v1.1 — Vals: 99%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: anthropic_claude-opus-5; Vals max effort

    Metric: ProofBench v1.1 overall percentage; Vals benchmark-specific harness

    Original source · Retrieved 2026-09-06

    Exact model and benchmark row; current source value agrees at sheet rounding. Uses published default parameters; source warns individual benchmarks may override them. Known mismatched rows/harnesses are not imported.

  74. Vibe Code Bench v1.1 — Vals: 88.4%

    Contributes to the score. Source category: independent. Review status: reviewed-independent.

    Configuration: anthropic_claude-opus-5; Vals max effort

    Metric: Vibe Code Bench v1.1 overall percentage; Vals benchmark-specific harness

    Original source · Retrieved 2026-09-06

    Exact model and benchmark row; current source value agrees at sheet rounding. Uses published default parameters; source warns individual benchmarks may override them. Known mismatched rows/harnesses are not imported.

  75. MineBench: 2106 native-points

    Not used in the score. Source category not recorded. Review status: excluded.

    Unresolved Pro/base model identities in source catalog; exclude until matched.

  76. OSWorld 2.0: 31.43%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  77. ARC-AGI-1: 97.5%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  78. ARC-AGI-3 (standard harness): 30.16%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  79. Terminal-Bench Science 0.1: 30%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Source selects the maximum score across efforts; exact highest-effort result is not established.

  80. SRE-Bench (pass@1): 12.5%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Source selects the maximum score across efforts; exact highest-effort result is not established.

  81. OSWorld 2.0 — OpenAI H2H (offline subset): 70.2%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Source selects the maximum score across efforts; exact highest-effort result is not established.

  82. Agents' Last Exam — OpenAI H2H: 55.5%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Source selects the maximum score across efforts; exact highest-effort result is not established.

  83. HealthBench Professional (length-adjusted): 56.4%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Source selects the maximum score across efforts; exact highest-effort result is not established.

  84. Harvey LAB-AA: 93.46%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  85. EnterpriseOps-Gym-AA: 47.48%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  86. MLCR-AA (Medical Long Context Reasoning): 55.56%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  87. AA-AnalystAgent: 53.75%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  88. DeepSearchQA: 95%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  89. Vending-Bench 2: 11181.87 native-points

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  90. ZeroBench: 26%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  91. Mystery Game Puzzles v1.0.4 — Epoch: 59%

    Not used in the score. Source category not recorded. Review status: unresolved.

    Inherited cell lacks a matched, reviewed source/configuration record.

  92. RuneBench (30m, mean ln(1 + XP/min)): 5.705729 native-points

    Contributes to the score. Source category: independent. Review status: reviewed-archive.

    Configuration: opus5-xhigh

    Metric: Arithmetic mean of ln(1 + peakXpRate) over all 16 skills, exactly as the owner’s Heatmap.js default ranking. Data rates are already normalized; do not divide by 200 again.

    Original source · Retrieved 2026-09-06

    Matched exact metric/value and the archived benchmark-specific configuration review. Unknown native effort stays explicitly unknown.

  93. VoxelBench (text-to-voxel, Glicko-2): 2315 native-points

    Contributes to the score. Source category: independent. Review status: reviewed-archive.

    Configuration: Claude Opus 5 (Max)

    Metric: Glicko-2 rating from the text-to-voxel board; not win percentage, deviation, image-input mode or MineBench Elo.

    Original source · Retrieved 2026-09-06

    Matched exact metric/value and the archived benchmark-specific configuration review. Unknown native effort stays explicitly unknown.

  94. Code Arena WebDev (overall): 1688 native-points

    Contributes to the score. Source category: independent. Review status: reviewed-archive.

    Configuration: claude-opus-5-max

    Metric: Published overall WebDev preference rating, not task success percentage or Arena Text score.

    Original source · Retrieved 2026-09-06

    Matched exact metric/value and the archived benchmark-specific configuration review. Unknown native effort stays explicitly unknown.

Cite this model record

The Latent. “Opus 5: score and benchmark evidence.” 2026-09-06. Release tli-2026-v5.1-2026-09-06.

Download release data and definitions (JSON) · Data usage terms