Sensitive Content Tagging on SenBen 1.0 (test)
64.7F1SBGemini 3 Pro†
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini 3 Pro†ms/fr=5579, $/2K=26.58, Reasoning Level=low reas.2026.04 | 64.7 | |
| Gemini 3 Flashms/fr=6121, $/2K=5.80, Reasoning Level=low reas.2026.04 | 58.3 | |
| Q2L-balParams=241M, ms/fr=733, VRAM=1.2 GB, $/2K=02026.04 | 42.8 | |
| Claude Opus 4.6ms/fr=4555, $/2K=20.022026.04 | 40.4 | |
| GLM-4.6VParams=10.3B, ms/fr=17056, VRAM=21.5 GB, $/2K=0, Reasoning Level=reas.2026.04 | 36.4 | |
| GPT-5.2ms/fr=9019, $/2K=16.25, Reasoning Level=med. reas.2026.04 | 36.2 | |
| Qwen3-VL-8BParams=8.3B, ms/fr=5,614, VRAM=18.8 GB, $/2K=02026.04 | 34 | |
| Claude Sonnet 4.6ms/fr=3438, $/2K=12.142026.04 | 33.9 | |
| GPT-5-minims/fr=13412, $/2K=4.49, Reasoning Level=med. reas.2026.04 | 33 |