DailySafe - Helpfulness Rate (HR): leaderboard

DailySafe (arXiv 2611.26298v1) dynamic multimodal safety benchmark (Helpfulness Rate (HR)) over 195 everyday multi-turn image-text cases.

Metric: Helpfulness Rate (%, 3-run mean over 195 multi-turn multimodal cases). Source: arxiv.org. Saturation forecast: Around February 2027. 16 models tracked. Updated 22 Nov 2026.

Top models

#ModelScoreOverall rank
1Claude Opus 5.594.53#9
2GPT-6 Sol94.53#33
3Gemini 3.8 Flash87.52#43
4GPT-5.6 Sol85.3#36
5Gemini 3.1 Pro (Preview)84.96#102
6Kimi K377.78#62
7Gemini 3.5 Flash76.07#110
8GLM-5.3 Flash69.57#137
9Qwen3.8-Flash-Next62.22#122
10Claude Sonnet 560.51#124
11MiMo-V2.6-Pro52.31#94
12Qwen 3.8 27B47.86#232
13DeepSeek V4.1 Flash44.44#69
14MiMo-V2.6-Flash36.07#183
15Qwen 3 VL 8B (Thinking)8.38#924

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

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