MMM-Bench - Level 1 (Business Function): leaderboard

MMM-Bench (2026, Alibaba) zero-shot document classification at taxonomy level 1, Business Function, for real enterprise documents given as OCR text and page images. Macro-F1 in percent; higher is better.

Metric: Macro-F1 (%) at taxonomy level 1 (Business Function), zero-shot hierarchical classification of the test split (20 percent of 5,990 real Alibaba business documents in 12 domains, mostly Chinese) from OCR text, page images and filename against a flattened five-level taxonomy, temperature 0.1; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 15 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3.1 Pro (Preview)86.1#99
2GPT-5.485.98#135
3Qwen 3 VL 235B A22B85#485
4Claude Opus 4.584.39#160
5Claude Opus 4.684.07#113
6Kimi K2.584.04#266
7GPT-5.283.27#185
8Qwen 3.6 Plus82.32#238
9Qwen 3.5 27B82.18#375
10MiniMax-M2.781.3#539
11Qwen 3.5 35B A3B80.33#492
12Qwen 3.5 9B79.8#828
13GLM-577.81#269
14Qwen 3 VL 8B69.87#917
15Qwen 2.5 VL 7B47.03#1862

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=mmm-bench-level-1-business-function · How It Works · Data refreshed daily, snapshot 2026-11-21.