AI BENCHY カテゴリ別失敗
反AIトリック
不正解
反AIトリック
不正解
反AIトリック で 不正解 が起きやすいAIモデルを確認し、弱点を早く見つけられます。 並び替え: 応答時間（平均） ↓.
関連する失敗理由
| 順位 | モデル | 企業 | 不正解 件数 | カテゴリスコア | 正解テスト | 応答時間（平均） |
|---|---|---|---|---|---|---|
| #34 | GPT-5 Nano medium | OpenAI | 1 | 7.0 | 2/3 | 37.7s |
| #18 | DeepSeek V3.2 medium | DeepSeek | 1 | 7.0 | 2/3 | 33.4s |
| #52 | GLM 4.7 Flash medium | Z.ai | 1 | 4.0 | 1/3 | 27.1s |
| #50 | Qwen3 Coder Next medium | Qwen | 2 | 1.3 | 0/3 | 15.3s |
| #46 | Kimi K2.5 none | Moonshot AI | 3 | 2.7 | 0/3 | 11.4s |
| #33 | DeepSeek V3.2 none | DeepSeek | 1 | 10.0 | 0/3 | 8.79s |
| #16 | Gemini 2.5 Flash medium | 1 | 7.3 | 2/3 | 6.98s | |
| #49 | GLM 4.7 Flash none | Z.ai | 3 | 10.0 | 0/3 | 6.59s |
| #48 | Qwen3 Coder Next none | Qwen | 1 | 2.3 | 0/3 | 4.39s |
| #45 | Trinity Large Preview none | Arcee AI | 3 | 10.0 | 0/3 | 3.59s |
| #31 | GLM 5 none | Z.ai | 2 | 4.0 | 1/3 | 3.39s |
| #29 | Qwen3.5 Plus 2026-02-15 none | Qwen | 2 | 4.0 | 1/3 | 2.74s |
| #17 | Gemini 3.1 Flash Lite Preview low | 1 | 7.0 | 2/3 | 2.18s | |
| #47 | GPT-4o-mini none | OpenAI | 2 | 4.0 | 1/3 | 1.83s |
| #42 | Qwen3.5-35B-A3B none | Qwen | 3 | 10.0 | 0/3 | 1.76s |
| #53 | Grok 4.1 Fast none | X AI | 2 | 1.3 | 0/3 | 1.73s |
| #37 | Qwen3.5-Flash none | Qwen | 3 | 2.3 | 0/3 | 1.62s |
| #20 | Gemini 3 Flash Preview none | 1 | 7.0 | 2/3 | 1.59s | |
| #44 | GPT-5.4 none | OpenAI | 3 | 10.0 | 0/3 | 1.41s |
| #54 | MiMo-V2-Flash none | Xiaomi | 3 | 10.0 | 0/3 | 1.36s |
| #22 | Gemini 3.1 Flash Lite Preview none | 1 | 6.0 | 1/3 | 1.16s | |
| #40 | Qwen3.5-122B-A10B none | Qwen | 2 | 4.0 | 1/3 | 927ms |
| #41 | Qwen3.5-27B none | Qwen | 2 | 4.0 | 1/3 | 796ms |
| #38 | Gemini 2.5 Flash none | 3 | 10.0 | 0/3 | 668ms | |
| #55 | LFM2-24B-A2B none | Liquid | 3 | 10.0 | 0/3 | 471ms |
| #51 | Mercury 2 none | Inception | 3 | 10.0 | 0/3 | 466ms |