Google DeepMind·released 2025-06-051 source
Gemini 2.5 Pro (Jun 2025) benchmark scores: 23 benchmarks tracked. 26% of its scores are independently reproduced — the rest are self-reported or unverified.
Leads on 0 of 23 benchmarks·best result 91.7 on Fiction.LiveBench (unverified)·6 of 23 independently reproduced
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| Fiction.LiveBench | 91.7accuracy (%) | unverifiedT1 | 2025-06-05Fiction.live leaderboard |
| OTIS Mock AIME 2024-2025 | 84.71accuracy (%) | reproduced· optimizedT1 | 2025-06-05Epoch AI |
| Lech Mazur Writing | 83.8accuracy (%) | unverifiedT1 | 2025-06-05lechmazur/writing Github repository |
| Aider polyglot | 83.1accuracy (%) | unverified· optimizedT1 | 2025-06-05https://aider.chat/docs/leaderboards/ |
| GPQA diamond | 80.39accuracy (%) | reproduced· optimizedT1 | 2025-06-05Epoch AI |
| DTBench | 70.67accuracy (%) | unverifiedT1 | 2025-06-05https://conceptualreasoning.ai/dtbench |
| SWE-Bench verified | 57.56accuracy (%) | reproduced· optimizedT1 | 2025-06-05Epoch AI |
| METR Time Horizons | 55.44accuracy (%) | unverifiedT1 | 2025-06-05METR - Measuring AI Ability to Complete Long Tasks |
| SimpleBench | 54.88accuracy (%) | unverifiedT1 | 2025-06-05https://lmcouncil.ai/benchmarks |
| WeirdML | 54.03accuracy (%) | unverifiedT1 | 2025-06-05WeirdML Leaderboard |
| DeepResearch Bench | 42.8accuracy (%) | unverifiedT1 | 2025-06-05DeepResearchBench Leaderboard |
| ARC-AGI | 41accuracy (%) | unverified· optimizedT1 | 2025-06-05ARC Prize Leaderboard |
| LMCA | 40.91accuracy (%) | unverifiedT1 | 2025-06-05https://conceptualreasoning.ai/lmca |
| Terminal Bench | 32.64accuracy (%) | unverifiedT1 | 2025-06-05https://www.tbench.ai/leaderboard/terminal-bench/2.0 |
| FrontierMath-Tiers-1-3-v2-Private | 24.56accuracy (%) | reproducedT1 | 2025-06-05Epoch AI |
| GDPval | 23.3accuracy (%) | unverifiedT2 | 2025-06-05 |
| VPCT | 19.6accuracy (%) | unverifiedT1 | 2025-06-05VPCT leaderboard |
| HLE | 17.69accuracy (%) | unverifiedT2 | 2025-06-05 |
| Chess Puzzles | 15.82accuracy (%) | reproducedT1 | 2025-06-05Epoch AI |
| ARC-AGI-2 | 4.86accuracy (%) | unverifiedT2 | 2025-06-05 |
| GSO-Bench | 3.92accuracy (%) | unverifiedT1 | 2025-06-05https://gso-bench.github.io/index.html |
| Remote Labor Index | 0.83accuracy (%) | unverifiedT2 | 2025-06-05 |
| FrontierMath-Tier-4-v2-Private | 0accuracy (%) | reproducedT1 | 2025-06-05Epoch AI |
Claims drawn from cited facts, not live model generation.
This model was developed by Google DeepMind. Its 2025 release makes it a recent vintage, meaning its design reflects the current state of the art rather than an earlier generation of methods.
2 cited facts
This model is scored on 23 tracked benchmarks, but it holds no current top score on any of them. Its record can be taken at face value: the results have been independently reproduced, meaning outside evaluation backs the scores up.
3 cited facts
Its mean gap to the best recorded score across its tracked benchmarks is 38.82 points.
1 cited fact
1 facts cross-checked across data sources: 1 single-source.
Source facts, citations, and refresh stamp for this record.
Sources: epoch_benchmarks
Gemini 2.5 Pro (Jun 2025) is an AI model developed by Google DeepMind, released 2025-06-05. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.
Gemini 2.5 Pro (Jun 2025) has recorded scores on 23 benchmarks, each shown with its evidence status.
Gemini 2.5 Pro (Jun 2025) has recorded scores on 23 benchmarks — Lech Mazur Writing, GPQA diamond, OTIS Mock AIME 2024-2025, DTBench, WeirdML, LMCA, and 17 more. The full table above shows each score with its evidence status.
6 of 23 recorded scores (26%) are independently reproduced rather than self-reported by the lab.
Gemini 2.5 Pro (Jun 2025) has 23 tracked claims: 6 independently reproduced, 17 unverified.