SynthDocBench - Complex Reasoning: leaderboard
Metric: Accuracy (%) on multi-hop complex-reasoning questions: GPT-5 judge scores each answer 0 to 10 against a deterministic reference and an answer counts when the score is at least 6; pages rasterized at 144 DPI in 5-page strips, vision-only input, temperature 0; higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 78.9 |
| 2 | Qwen 3 VL 235B A22B | 61.1 |
| 3 | GPT-5.4 | 45.7 |
| 4 | InternVL3-78B | 39.7 |
| 5 | GPT-4o | 36 |
| 6 | Claude Sonnet 4.5 | 33.7 |
| 7 | Qwen 2.5 VL 7B Instruct | 1.2 |
Interactive version: theaggregate.ai/benchmark?slug=synthdocbench-complex-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-29.