SafeBench (Multimodal) — leaderboard

Safety evaluation framework and leaderboard for multimodal LLMs, measuring safety across attacks, risk categories, and vision-language prompt formats.

Source: safebench-mm.github.io.

Interactive version: theaggregate.ai/benchmark?slug=safebench-multimodal · How the rankings work · Data refreshed daily, snapshot 2026-07-22.