Atlas

Benchmarks

← All benchmarks

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)

ModelScore
Gemini 3 Flash Preview96.9
Gemini 3.1 Pro Preview94.8
Gemini 3.5 Flash94.8
Gemini 3.6 Flash94.8
Claude Fable 592.8
Gemini 3.5 Flash-Lite90.7
Sonnet 589.7
Opus 4.887.6
GPT-5.587.6
Muse Spark 1.186.6
Qwen3-VL-235B-A22B-Instruct86.6
Gemini 2.5 Pro84.5
Kimi K384.5
Qwen3.7-Plus83.5
GPT-5.4 Mini82.5
GPT-5.6 Sol82.5
GLM 5V Turbo81.4
GPT-5.6 Luna81.4
GPT-5.6 Terra79.4
Kimi K2.678.3
Qwen3.5-27B78.3
Loading Atlas data…