Acropolis

Activity stream

Feed

Source snapshot refreshes, derived benchmark leader observations, and public/private licensing gate notices. Build 17 Nov 2026.

Subscribe via RSS
  1. EvaluationEpoch AI Benchmarking Hub

    GPT-6 Astra leads ARC-AGI-2

    95% on Score (Epoch AI Benchmarking Hub, 299 evaluated rows).

  2. EvaluationCursorBench

    Gemini 4 Carbon leads CursorBench

    60.4% on Score (CursorBench, 91 evaluated rows).

  3. EvaluationEpoch AI Benchmarking Hub

    Opus 5.5 leads FrontierCode

    54.6% on Main score (Epoch AI Benchmarking Hub, 51 evaluated rows).

  4. EvaluationEpoch AI Benchmarking Hub

    GPT-5.2 leads GDPval

    49.7% on Win rate (Epoch AI Benchmarking Hub, 11 evaluated rows).

  5. EvaluationEpoch AI Benchmarking Hub

    GPT-6 Astra leads HLE

    54.8% on Accuracy (Epoch AI Benchmarking Hub, 56 evaluated rows).

  6. EvaluationEpoch AI Benchmarking Hub

    Sonnet 4.6 leads OSWorld

    72.1% on Score (Epoch AI Benchmarking Hub, 38 evaluated rows).

  7. EvaluationEpoch AI Benchmarking Hub

    Gemini 4 Carbon leads Vending-Bench 2

    $16,484.21 on Score (Epoch AI Benchmarking Hub, 72 evaluated rows).

  8. Source updateCursorBench

    CursorBench snapshot updated

    364 results ingested.

  9. Source updateEpoch AI Benchmarking Hub

    Epoch AI Benchmarking Hub snapshot updated

    1893 results ingested.

  10. Source updateLMArena

    LMArena snapshot updated

    26565 results ingested.

  11. Source updateVals AI benchmarks

    Vals AI benchmarks snapshot updated

    1195 results ingested.

  12. EvaluationLMArena

    Fable 5.2 leads LMArena Agent

    0.3 on Agent score (LMArena, 64 evaluated rows).

  13. EvaluationLMArena

    Fable 5.2 leads LMArena WebDev

    1,826 on Arena rating (LMArena, 156 evaluated rows).

  14. EvaluationMETR time horizons

    Claude Mythos Preview (early) leads METR time horizon

    1,044.8 min on 50% time horizon (METR time horizons, 23 evaluated rows).

  15. EvaluationVals AI benchmarks

    Sonnet 5.5 leads Code Migration

    87.4% on Accuracy (Vals AI benchmarks, 38 evaluated rows).

  16. EvaluationVals AI benchmarks

    Fable 5.2 leads EMB

    77.4% on Accuracy (Vals AI benchmarks, 38 evaluated rows).

  17. EvaluationVals AI benchmarks

    Fable 5.2 leads Finance Agent

    69.5% on Accuracy (Vals AI benchmarks, 38 evaluated rows).

  18. EvaluationVals AI benchmarks

    Fable 5.2 leads Vals Index

    71.3% on Accuracy (Vals AI benchmarks, 38 evaluated rows).

  19. EvaluationVals AI benchmarks

    Gemini 4 Argon leads Legal Agent Benchmark (HLAB-Vals)

    27.1% on Accuracy (Vals AI benchmarks, 38 evaluated rows).

  20. EvaluationVals AI benchmarks

    Gemini 4 Carbon leads Legal Research

    55.8% on Accuracy (Vals AI benchmarks, 38 evaluated rows).

  21. EvaluationVals AI benchmarks

    Gemini 4 Hydrogen leads Tax Agent

    52.3% on Accuracy (Vals AI benchmarks, 38 evaluated rows).

  22. EvaluationVals AI benchmarks

    Fable 5.2 leads Terminal-Bench 4

    65.2% on Accuracy (Vals AI benchmarks, 38 evaluated rows).

  23. EvaluationVals AI benchmarks

    Fable 5.2 leads Vibe Code Bench

    93.8% on Accuracy (Vals AI benchmarks, 38 evaluated rows).

  24. Source updateArtificial Analysis

    Artificial Analysis snapshot updated

    Metadata synced (link-only in public mode).

  25. Source updateMETR time horizons

    METR time horizons snapshot updated

    78 results ingested.

  26. Source updateBug Hunt Bench

    Bug Hunt Bench snapshot updated

    1966 results ingested.

  27. EvaluationLMArena

    GPT Image 2.5 Sunburst leads LMArena Image Edit

    1,519 on Arena rating (LMArena, 48 evaluated rows).

  28. EvaluationLMArena

    GPT Image 2.5 Sunburst leads LMArena Text-to-Image

    1,429 on Arena rating (LMArena, 85 evaluated rows).

  29. EvaluationLMArena

    Gemini 4 Carbon leads LMArena Text (style control)

    1,526 on Arena rating (LMArena, 421 evaluated rows).

  30. EvaluationLMArena

    Gemini 4 Carbon leads LMArena Text

    1,538 on Arena rating (LMArena, 421 evaluated rows).

  31. EvaluationBug Hunt Bench

    Fable 5.2 leads Bug Hunt Bench

    58 on Bugs fixed (Bug Hunt Bench, 145 evaluated rows).

  32. EvaluationLMArena

    Fable 5 leads LMArena Document (style control)

    1,500 on Arena rating (LMArena, 46 evaluated rows).

  33. EvaluationLMArena

    Fable 5.1 leads LMArena Document

    1,515 on Arena rating (LMArena, 46 evaluated rows).

  34. EvaluationLMArena

    Wan 2.5 Preview leads LMArena Image-to-Video

    1,322 on Arena rating (LMArena, 2 evaluated rows).

  35. EvaluationLMArena

    GPT-5.6 Sol leads LMArena Search (style control)

    1,255 on Arena rating (LMArena, 42 evaluated rows).

  36. EvaluationLMArena

    Qwen3.8-Max leads LMArena Search

    1,260 on Arena rating (LMArena, 42 evaluated rows).

  37. EvaluationLMArena

    Wan 2.6 Image leads LMArena Text-to-Video

    1,324 on Arena rating (LMArena, 2 evaluated rows).

  38. EvaluationLMArena

    Fable 5 leads LMArena Vision (style control)

    1,309 on Arena rating (LMArena, 167 evaluated rows).

  39. EvaluationLMArena

    Fable 5 leads LMArena Vision

    1,326 on Arena rating (LMArena, 167 evaluated rows).

  40. EvaluationEpoch AI Benchmarking Hub

    Gemini 4 Carbon leads GPQA diamond

    96.5% on Accuracy (best score across scorers) (Epoch AI Benchmarking Hub, 338 evaluated rows).

  41. ChangelogCursorBench

    CursorBench release note

    Reporting: Updated Sonnet 5.5 results to account for adjusted pricing.

  42. ChangelogCursorBench

    CursorBench release note

    Tasks: CursorBench 4.0

  43. ChangelogCursorBench

    CursorBench release note

    Reporting: Updated Sonnet 5 results to account for adjusted pricing.

  44. ChangelogCursorBench

    CursorBench release note

    Reporting: Updated GPT-5.6 Terra and Luna results to account for adjusted pricing.

  45. ChangelogCursorBench

    CursorBench release note

    Reporting: Updated GPT-5.6 Sol, Terra, and Luna results to account for cache write costs.

  46. ChangelogCursorBench

    CursorBench release note

    Tasks: CursorBench 3.2

  47. ChangelogCursorBench

    CursorBench release note

    Tasks: CursorBench 3.1

  48. EvaluationEpoch AI Benchmarking Hub

    GPT-5.5 Pro Pre-release leads FrontierMath-2025-02-28-Private

    52.4% on Accuracy (best score across scorers) (Epoch AI Benchmarking Hub, 101 evaluated rows).

  49. EvaluationEpoch AI Benchmarking Hub

    GPT-5.5 leads Terminal Bench

    84.7% on Accuracy (Epoch AI Benchmarking Hub, 202 evaluated rows).

  50. EvaluationEpoch AI Benchmarking Hub

    Opus 4.7 leads SWE-Bench verified

    83.5% on Accuracy (best score across scorers) (Epoch AI Benchmarking Hub, 35 evaluated rows).

  51. ChangelogCursorBench

    CursorBench release note

    Tasks: CursorBench 3.0