Kegagalan kategori AI BENCHY
Trik anti-AI: Tidak mengikuti instruksi
Trik anti-AI
Tidak mengikuti instruksi
Lihat model AI mana yang paling mungkin mengalami Tidak mengikuti instruksi di Trik anti-AI, agar Anda bisa menemukan titik lemahnya lebih cepat.
Alasan kegagalan
30/30
Filter model
Tidak ada model yang cocok dengan pencarian dan filter saat ini.
| Peringkat | Model | Perusahaan | Jumlah Tidak mengikuti instruksi | Skor kategori | Total Biaya | Tes benar | Waktu respons (rata-rata) |
|---|---|---|---|---|---|---|---|
| #186 | Hy3 preview none | Tencent | 2 | 4.8 | $0.003 | 1/4 | 11.1s |
| #19 | GPT-5 Mini medium | OpenAI | 1 | 7.1 | $0.159 | 2/4 | 13.9s |
| #23 | Step 3.7 Flash medium | Stepfun | 1 | 8.7 | $0.376 | 3/4 | 9.65s |
| #25 | GPT-5.2 medium | OpenAI | 1 | 6.5 | $0.548 | 2/4 | 7.81s |
| #42 | Gemini 3.1 Flash Lite Preview medium | 1 | 9.1 | $0.068 | 3/4 | 2.33s | |
| #44 | Gemini 3.1 Flash Lite medium | 1 | 9.1 | $0.071 | 3/4 | 2.39s | |
| #45 | Kimi K2.6 medium | Moonshot AI | 1 | 7.0 | $0.888 | 2/4 | 11.6s |
| #54 | MiniMax M3 medium | Minimax | 1 | 5.5 | $0.131 | 1/4 | 14.9s |
| #59 | Mercury 2 medium | Inception | 1 | 6.9 | $0.058 | 2/4 | 1.12s |
| #61 | GPT-5.3 Chat none | OpenAI | 1 | 6.7 | $0.433 | 2/4 | 3.86s |
| #72 | GLM 5V Turbo medium | Z.ai | 1 | 7.2 | $0.457 | 2/4 | 10.8s |
| #74 | DeepSeek V4 Pro none | DeepSeek | 1 | 3.2 | $0.034 | 0/4 | 4.02s |
| #96 | gpt-oss-120b medium | OpenAI | 1 | 6.7 | $0.011 | 2/4 | 10.2s |
| #102 | Gemini 3.1 Flash Lite Preview none | 1 | 7.5 | $0.018 | 2/4 | 1.04s | |
| #134 | Owl Alpha none | Openrouter | 1 | 3.4 | $0.000 | 0/4 | 2.78s |