Phare Harmful Misguidance (Spanish)
Safety · 2025-03-27
Phare task testing resistance to harmful or unsafe guidance. This row is the Spanish language split.
Top models (higher is better)
| Model | Score |
|---|---|
| Haiku 4.5 | 100.0 |
| Opus 4.6 | 100.0 |
| Sonnet 4.6 | 99.8 |
| Sonnet 5 | 99.8 |
| GPT-5.5 | 98.8 |
| Sonnet 4.5 | 98.2 |
| GPT-5 Mini | 98.2 |
| Opus 4.5 | 96.8 |
| Gemma 4 31B IT | 96.8 |
| Kimi K2.5 | 96.8 |
| Kimi K2.6 | 96.8 |
| Opus 4.1 | 96.5 |
| GPT-5 | 96.3 |
| GPT-5 Nano | 96.3 |
| Gemini 3.1 Flash-Lite | 96.2 |
| Gemini 1.5 Pro | 96.1 |
| Gemini 3.5 Flash | 95.9 |
| GPT-5.2 | 95.9 |
| Qwen3.7 Max Preview | 95.5 |
| Claude 3.7 Sonnet | 95.5 |
| Gemini 3.1 Pro Preview | 95.3 |
| GPT-5.1 | 95.3 |
| Claude 3.5 Sonnet (Oct 2024) | 95.1 |
| Haiku 3.5 | 94.7 |
| Qwen3 Max (rolling alias) | 94.7 |
| DeepSeek-V4-Pro | 94.1 |
| Gemini 2.5 Flash | 93.9 |
| DeepSeek-V4-Flash | 93.7 |
| Qwen-Plus (2025-01-25) | 93.7 |
| DeepSeek-R1-0528 | 93.5 |
| Gemini 3 Pro Preview | 93.5 |
| Grok 4.3 | 93.3 |
| GLM-5.2 | 93.1 |
| Qwen3.7-Plus | 92.9 |
| Gemini 2.0 Flash | 92.7 |
| DeepSeek-V3.1 | 92.1 |
| DeepSeek-V3-0324 | 91.9 |
| GPT-4o | 91.5 |
| gpt-oss-120b | 91.3 |
| Mistral Medium 3.1 | 91.1 |