SeePhys Pro: leaderboard
We introduce SeePhys Pro, a fine-grained modality transfer benchmark that studies whether models preserve the same reasoning capability when critical information is progressively transferred from text to image. Unlike standard vision-essential benchmarks that evaluate a single input form, SeePhys Pro features four semantically aligned variants for each problem with progressively increasing visual
Metric: Cons4 (self-reported). Source: benchmarklist.com. Saturation forecast: Around August 2027. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 47 | #58 |
| 2 | Gemini 3 Pro | 43 | #80 |
| 3 | Claude Opus 4.7 | 33.5 | #49 |
| 4 | GPT-5.4 | 32.6 | #79 |
| 5 | Qwen 3.6 Flash | 29.9 | #214 |
| 6 | Kimi K2.5 | 26 | #144 |
| 7 | Qwen 3.5 122B A10B | 25.5 | #173 |
| 8 | Claude Opus 4.6 | 19 | #63 |
| 9 | Qwen 3.5 9B | 12.8 | #379 |
| 10 | Qwen 3.5 27B | 9.9 | #187 |
| 11 | Gemma 4 26B A4B (IT) | 9 | #297 |
| 12 | GPT-5 | 8.9 | #98 |
| 13 | Gemma 4 31B (IT) | 8.9 | #176 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=seephys-pro · How It Works · Data refreshed daily, snapshot 2026-10-20.