Roboflow Vision Evals - Data Extraction
Multimodal · 2026-07-22
The Roboflow Vision Evals data-extraction task measures exact-match accuracy on 97 image samples that require extracting structured information.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 3 Flash Preview | 96.9 |
| Gemini 3.1 Pro Preview | 94.8 |
| Gemini 3.5 Flash | 94.8 |
| Gemini 3.6 Flash | 94.8 |
| Claude Fable 5 | 92.8 |
| Gemini 3.5 Flash-Lite | 90.7 |
| Sonnet 5 | 89.7 |
| Opus 4.8 | 87.6 |
| GPT-5.5 | 87.6 |
| Muse Spark 1.1 | 86.6 |
| Qwen3-VL-235B-A22B-Instruct | 86.6 |
| Gemini 2.5 Pro | 84.5 |
| Kimi K3 | 84.5 |
| Qwen3.7-Plus | 83.5 |
| GPT-5.4 Mini | 82.5 |
| GPT-5.6 Sol | 82.5 |
| GLM 5V Turbo | 81.4 |
| GPT-5.6 Luna | 81.4 |
| GPT-5.6 Terra | 79.4 |
| Kimi K2.6 | 78.3 |
| Qwen3.5-27B | 78.3 |