DistortBench - Color: leaderboard

DistortBench (2026) shows a vision-language model a distorted photo without its original and asks it to name the distortion type and severity among four options (chance 25 percent). Joint accuracy on the five color distortions (diffusion, shift, quantization, saturation up and down), in percent; higher is better.

Metric: Joint accuracy (%) on the five color distortions (diffusion, shift, quantization, saturation up and down), a category of DistortBench's 13,500 four-choice questions (27 distortion types at five KADID-10k calibrated severity levels on 100 KADIS-700k photos, no reference image), the answer naming both type and severity, answered-only accuracy (unparseable responses, at most 3.4 percent, leave the denominator), chance 25; higher is better. Source: arxiv.org. Saturation forecast: Around December 2027. 14 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Qwen 3.5 27BNon-reasoning58.3#380
2InternVL3.5-38B57.3#1026
3InternVL3.5-14B56.4#1519
4Qwen 3.5 4B53.5#1270
5Qwen 3.5 9B52.8#838
6Kimi-VL-A3B-Instruct52.1#2272
7Gemma 3 27B51.1#1672
8Qwen 3.5 35B A3BNon-reasoning51.1#494
9Gemma 3 12B50.8#2091
10GPT-5.450#138
11InternVL3.5-8B48.4#1618
12Kimi-VL-A3B (Thinking)42.4
13Qwen 2 VL 7B36.3#2300
14Gemma 3 4B26.4#5347

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=distortbench-color · How It Works · Data refreshed daily, snapshot 2026-11-22.