Uni-SafeBench - Unsafe T2I Generation: 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 requests for unsafe images; higher is better.

Metric: Both-Safe rate (%) on the 390 unsafe text-to-image generation queries of Uni-SafeBench (requests for unsafe images), 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: Rough model projection: around 2026. 9 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Janus-flow76.2
2Harmon-1.5B72.3
3Emu370.8
4Tar-1.5B68.7
5Gemini 2.5 Flash Image66.9
6Janus-Pro-7B66.2#6889
7Show-o2-7B65.9
8BAGEL-7B-MoT63.6#4668
9Tar-7B58.2

Interactive version: theaggregate.ai/benchmark?slug=uni-safebench-unsafe-t2i-generation · How It Works · Data refreshed daily, snapshot 2026-11-22.