RLTasksby ulam.ai
Models / Google

Google / MODEL EVIDENCE

Gemini 2.5 Pro Preview (Jun 2025)

19 benchmarks · 29 observations · 18 sources

Scores at a glance

19 benchmarks · large dot: Gemini 2.5 Pro Preview (Jun 2025); small dots: closest models where reported

GPQA Diamond ↗84.85%
84.8584.85

Gemini 2.5 Pro Preview (Jun 2025) · GPQA diamond · Best score (across scorers) · % ↗

1 more reported result

66.67% · Gemini 2.5 Pro Preview (Jun 2025) · GPQA diamond · Best score (across scorers) · % · Epoch AI · GPQA diamond ↗

MATH Level 5 ↗95.9%
3.2998.13

Gemini 2.5 Pro Preview (Jun 2025) · MATH level 5 · Best score (across scorers) · % ↗

ARC-AGI-1 ↗31.3%
098.5

Gemini 2.5 Pro Preview (Jun 2025) · ARC-AGI · Score · % ↗

1 more reported result

16% · Gemini 2.5 Pro Preview (Jun 2025) · ARC-AGI · Score · % · Epoch AI · ARC-AGI ↗

GeoBench ↗86%
3286

Gemini 2.5 Pro Preview (Jun 2025) · GeoBench · ACW Country % · % ↗

1 more reported result

81% · Gemini 2.5 Pro Preview (Jun 2025) · GeoBench · ACW Country % · % · Epoch AI · GeoBench ↗

GSO ↗3.92%
1.344.12

Gemini 2.5 Pro Preview (Jun 2025) · GSO-Bench · Score OPT@1 · % ↗

SimpleBench ↗62.4%
45.880.6

Gemini 2.5 Pro Preview (Jun 2025) · SimpleBench · Score (AVG@5) · % ↗

VPCT ↗46.4%
46.446.4

Gemini 2.5 Pro Preview (Jun 2025) · VPCT · Correct · % ↗

1 more reported result

40.5% · Gemini 2.5 Pro Preview (Jun 2025) · VPCT · Correct · % · Epoch AI · VPCT ↗

ARC-AGI-2 ↗0%
095

Gemini 2.5 Pro Preview (Jun 2025) · ARC-AGI-2 · Score · % ↗

EnigmaEval ↗5.57%
5.575.57

Gemini 2.5 Pro Preview (Jun 2025) · EnigmaEval · Accuracy · % ↗

1 more reported result

2.36% · Gemini 2.5 Pro Preview (Jun 2025) · EnigmaEval · Accuracy · % · Epoch AI · EnigmaEval ↗

Closest models by score

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

o3 →OpenAI
3.4 percentile-point gap

3 of 3 shared benchmark comparisons

6.2 percentile-point gap

3 of 3 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 →

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. FutureSearch Evals ↗

    FutureSearch

  4. ARC-AGI-2 ↗

    Epoch AI

  5. FrontierMath Tier 4 (v1, old) ↗

    Epoch AI

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

    Epoch AI

  7. GPQA Diamond ↗

    Epoch AI

  8. Humanity's Last Exam ↗

    Epoch AI

  9. MATH Level 5 ↗

    Epoch AI

  10. Remote Labor Index ↗

    Epoch AI

  11. The Agent Company ↗

    Epoch AI

  12. Capabilities & benchmarking — retained external evaluation records ↗

    Epoch AI · 2026

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

    Fiction.live · 2025

  14. GeoBench — GeoGuessr model benchmark ↗

    GeoBench

  15. Writing benchmark — incorporating mandatory story elements ↗

    lechmazur

  16. EnigmaEval ↗

    Scale Labs / SEAL

  17. AI Model Benchmarks — LM Council ↗

    LM Council

  18. Measuring AI Ability to Complete Long Software Tasks ↗

    METR

Release dates reported by sources

2025-05-06 · 2025-06-05

Names used in source tables

Gemini 2.5 Pro Preview (Jun 2025)