COHERENCE - Partial Match: leaderboard
Metric: Partial-match score (Kendall's tau between predicted and true order, rescaled to 0-100), over all 6,161 interleaved image-text documents whose images were removed: the model assigns each candidate image to its placeholder; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 19 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 90.11 |
| 2 | Claude Sonnet 4.6 (Thinking) | 88.76 |
| 3 | Qwen 3.5 397B A17B | 88.37 |
| 4 | Kimi K2.5 | 86.74 |
| 5 | GPT-5.4 (High) | 86.54 |
| 6 | Qwen 3.5 122B A10B | 86.52 |
| 7 | Qwen 3.5 35B A3B | 85.06 |
| 8 | Doubao-Seed-2.0-Pro-260215 | 84.76 |
| 9 | Qwen 3 VL 235B A22B | 81.61 |
| 10 | Qwen 3.5 4B | 81.16 |
| 11 | GLM-4.6V | 79.29 |
| 12 | Step3 VL 10B | 76.75 |
| 13 | Qwen 3 VL 8B Instruct | 75.86 |
| 14 | GPT-5.2 | 74.84 |
| 15 | Qwen 3 VL 4B Instruct | 72.9 |
Interactive version: theaggregate.ai/benchmark?slug=coherence-partial-match · How It Works · Data refreshed daily, snapshot 2026-10-07.