XL-DocBench (MDocAgent): leaderboard
XL-DocBench run through the MDocAgent multi-agent document pipeline. Its best model lands 9 points below the same model under SimpleDoc, which is the benchmark’s own answer to whether agents are automatically better.
Metric: Accuracy (%). Source: officeintelligence.github.io. Saturation forecast: Around March 2028. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 35 | #79 |
| 2 | Claude Opus 4.6 | 33.2 | #63 |
| 3 | GPT-5.2 | 33 | #110 |
| 4 | Kimi K2.5 | 31.1 | #144 |
| 5 | DeepSeek V3.2 | 24 | #205 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=xl-docbench-mdocagent · How It Works · Data refreshed daily, snapshot 2026-10-20.