FlagEval VQA - Text Understanding (Chinese): leaderboard
BAAI FlagEval’s multimodal VQA leaderboard, Text Understanding (Chinese) slice: reading and reasoning over Chinese text inside images.
Metric: Score. Source: flageval.baai.ac.cn. Saturation forecast: Not forecast. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | ERNIE-4.5-VL-424B-A47B | 80.3 | #391 |
| 2 | Gemini 2.5 Pro | 78.6 | #152 |
| 3 | GLM-4.5V | 75.3 | #346 |
| 4 | Gemini 2.5 Flash (Thinking) | 75.2 | #245 (Gemini 2.5 Flash) |
| 5 | Step-3 | 73 | #365 |
| 6 | Gemini 2.5 Flash (Non-reasoning) | 71.4 | #245 (Gemini 2.5 Flash) |
| 7 | GPT-5 (High) | 66.2 | #98 (GPT-5) |
| 8 | GPT-5 (Low) | 65.8 | #98 (GPT-5) |
| 9 | Qwen 2.5 VL 7B | 65 | #643 |
| 10 | GPT-4.1 | 64.1 | #248 |
| 11 | O4 Mini (Low) | 63.2 | #180 (O4 Mini) |
| 12 | O4 Mini (High) | 62.8 | #180 (O4 Mini) |
| 13 | O3 Mini (Low) | 59.4 | #272 (O3 Mini) |
| 14 | O3 Mini (High) | 59 | #272 (O3 Mini) |
| 15 | Claude Sonnet 4 (Thinking) | 56.2 | #199 (Claude Sonnet 4) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=flageval-vqa-text-understanding-chinese · How It Works · Data refreshed daily, snapshot 2026-10-20.