Agent Evaluation on Auto-ClawEval Mini (104 environments)
94.2Safety ScoreMiniMax M2.7
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| MiniMax M2.7Family=MiniMax2026.04 | 94.2 | 35.7 | 100 | 44.9 | |
| GPT-5.4Family=OpenAI2026.04 | 93.3 | 51.2 | 100 | 56.5 | |
| GPT-5-nanoFamily=OpenAI2026.04 | 93.3 | 49.6 | 100 | 55.7 | |
| MiniMax M2.5Family=MiniMax2026.04 | 92.3 | 45 | 100 | 51.4 | |
| Claude Sonnet 4.6Family=Anthropic2026.04 | 90.4 | 50.6 | 100 | 54.2 | |
| GLM 5Family=Zipu AI2026.04 | 90.4 | 46.4 | 100 | 51.3 | |
| GLM 5 TurboFamily=Zipu AI2026.04 | 88.5 | 47.2 | 100 | 50.3 | |
| Claude Opus 4.6Family=Anthropic2026.04 | 87.5 | 49.3 | 100 | 52.1 |