CFMME - Application QA: leaderboard

CFMME (2026) is a Chinese financial multimodal evaluation set for vision-language models, built from financial documents, charts, seals, tables and exam questions. This board is accuracy in percent on questions about real financial document images; higher is better.

Metric: Accuracy (%) on question answering over real financial document images (application assessment) on CFMME (6,052 Chinese financial multimodal instances over eight financial image types), zero-shot, mean of three runs; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 14 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Qwen 3 VL 235B A22B (Thinking)54.94#459
2Qwen 3 VL 235B A22B Instruct52.57#532
3GLM-4.5VThinking51.35#754
4InternVL3.5-241B-A28BThinking50.08#733
5ERNIE 4.5 VL 28B A3B (Thinking)45.99
6Qwen 3 VL 8B (Thinking)44.22#910
7Qwen 3 VL 8B Instruct43.8#1029
8InternVL3.5-8BThinking43.29#1604
9ERNIE 4.5 VL 28B A3B42.07#2121
10GLM-4.1V-9B (Thinking)41.81
11GPT-4o37.05#751
12Kimi-VL-A3B (Thinking)36.37
13Kimi-VL-A3B-Instruct19.41#2253
14Llama 3.2 11B Instruct18.61#5174

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

Interactive version: theaggregate.ai/benchmark?slug=cfmme-application-qa · How It Works · Data refreshed daily, snapshot 2026-11-21.