Gemini 2.5 Pro (Preview 05-06): benchmark results
May 6 Gemini 2.5 Pro preview snapshot. Provider: Google. Released 2025-05-06. Access: API.
Aggregate Rating 1731 ± 9, rank #147 of 1656 rated models, from 261 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AVI-Bench | 57.21 | Overall score (%): mean of the four stage averages (percepti | 100 |
| AVI-Bench - Audio Multi-Instance Classification | 43.01 | Task score (%) on audio multi-instance classification (518 c | 100 |
| AVI-Bench - Audio Sensation QA | 29.48 | Task score (%) on primitive audio sensation (502 multiple-ch | 100 |
| AVI-Bench - Audio-Referenced Visual Retrieval | 74.42 | Task score (%) on audio-referenced visual retrieval (264 que | 100 |
| AVI-Bench - Audio-Visual Language Grounding | 35.08 | Task score (%) on audio-visual language grounding (503 video | 100 |
| AVI-Bench - Audio-Visual Question Answering | 73.93 | Task score (%) on audio-visual question answering (469 binar | 100 |
| AVI-Bench - Perception | 54.58 | Stage score (%) for the perception stage: mean of the audio | 100 |
| AVI-Bench - Primitive Sensation | 36.22 | Stage score (%) for the primitive sensation extension (AVI-B | 100 |
| AVI-Bench - Reasoning | 69.06 | Stage score (%) for the reasoning stage: mean of the two hal | 100 |
| AVI-Bench - Understanding | 68.97 | Stage score (%) for the understanding stage: mean of the two | 100 |
| AVI-Bench - Visual Multi-Instance Classification | 59.39 | Task score (%) on visual multi-instance classification (521 | 100 |
| AVI-Bench - Visual Sensation QA | 62.67 | Task score (%) on primitive visual sensation (1,200 multiple | 100 |
Interactive version: theaggregate.ai/model?slug=gemini-2-5-pro-preview-05-06 · How It Works · Data refreshed daily, snapshot 2026-10-20.