C3-Bench - Reversibility: leaderboard
Metric: Reversibility (%): share of C3-Bench image pairs for which a GPT-5.2 judge finds that the model's before-to-after and after-to-before captions describe the same changes in opposite directions; mean of three judge runs; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 26 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro | 73 |
| 2 | GPT-5.1 | 63 |
| 3 | GPT-5.2 | 62 |
| 4 | Gemini 3 Flash | 62 |
| 5 | Gemini 2.5 Pro | 52 |
| 6 | GPT-4o | 49 |
| 7 | Gemini 2.5 Flash | 48 |
| 8 | Qwen 3 VL 32B Instruct | 47 |
| 9 | InternVL3-38B | 46 |
| 10 | GPT-5 Mini | 45 |
| 11 | InternVL3-78B | 45 |
| 12 | Llama 4 Scout Instruct | 44 |
| 13 | Qwen 3 VL 8B Instruct | 40 |
| 14 | Qwen 2.5 VL 32B Instruct | 40 |
| 15 | Qwen 2.5 VL 72B Instruct | 38 |
Interactive version: theaggregate.ai/benchmark?slug=c3-bench-reversibility · How It Works · Data refreshed daily, snapshot 2026-09-29.