Uni-SafeBench - Image-Unsafe VQA: leaderboard

Uni-SafeBench (2026) tests the safety of unified multimodal models, vision-language models and LLMs on malicious requests across understanding and generation tasks, with a judge that checks both the context of each response and its intrinsic content. This board is the Both-Safe rate in percent on visual questions with a malicious image and a benign text; higher is better.

Metric: Both-Safe rate (%) on the 467 image-unsafe visual question answering queries of Uni-SafeBench (malicious image, benign text), judged by the Uni-Judger framework (malicious-intent extractor plus safety detector on Qwen3-VL-8B-Instruct), a response counts as safe only when it is safe or a refusal in the context of the extracted intent and its own content is not intrinsically unsafe, mean over repeated judge runs; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 14 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Chameleon (Uni-SafeBench checkpoint unspecified)98.9
2GPT-4o96.1#751
3Gemini 2.5 Flash Image92.1
4Qwen 2.5 VL 7B Instruct89.7#1902
5DeepSeek-VL-7B-Chat83.3#3756
6Tar-7B67.9
7Show-o2-7B60.4
8Harmon-1.5B60.2
9Tar-1.5B43.9
10BAGEL-7B-MoT40#4619
11DeepSeek-VL-7B-Base33.4
12Janus-Pro-7B32.3#6857
13Emu322.1
14Janus-flow19.3

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=uni-safebench-image-unsafe-vqa · How It Works · Data refreshed daily, snapshot 2026-11-21.