FlagEval VQA - Subject Knowledge: leaderboard
BAAI FlagEval’s multimodal VQA leaderboard, Subject Knowledge slice: questions needing domain knowledge on top of what the image shows.
Metric: Score. Source: flageval.baai.ac.cn. Saturation forecast: Estimated already saturated. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 77.7 | #152 |
| 2 | GPT-5 (Low) | 71.6 | #98 (GPT-5) |
| 3 | GPT-5 (High) | 68.9 | #98 (GPT-5) |
| 4 | Claude Sonnet 4 (Thinking) | 68.2 | #199 (Claude Sonnet 4) |
| 5 | Gemini 2.5 Flash (Thinking) | 67.6 | #245 (Gemini 2.5 Flash) |
| 6 | O3 Mini (High) | 65.5 | #272 (O3 Mini) |
| 7 | O4 Mini (High) | 64.2 | #180 (O4 Mini) |
| 8 | O3 Mini (Low) | 63.5 | #272 (O3 Mini) |
| 9 | Gemini 2.5 Flash (Non-reasoning) | 61.5 | #245 (Gemini 2.5 Flash) |
| 10 | Claude Sonnet 4 (Non-reasoning) | 58.1 | #199 (Claude Sonnet 4) |
| 11 | O4 Mini (Low) | 57.4 | #180 (O4 Mini) |
| 12 | Step-3 | 55.9 | #365 |
| 13 | ERNIE-4.5-VL-424B-A47B | 50 | #391 |
| 14 | GPT-4.1 | 49.3 | #248 |
| 15 | Mistral Medium 3.1 | 48 | #408 |
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-subject-knowledge · How It Works · Data refreshed daily, snapshot 2026-10-20.