HyperAI
Command Palette
Search for a command to run...
𝜏³-Banking
𝜏-Knowledgeフレームワークによるフィンテック顧客サポートベンチマーク。エージェントが大規模で非構造化な知識ベースを探索し、多段階のツール呼び出しを実行して現実的な銀行業務ワークフローを解決できるかをテストする。
- モデル
- 89 現在、 210 合計
- 中央値
- 17.5%
- 範囲
- 2.9%–50.3%
| ランク | 比較する | モデル | スコア |
|---|---|---|---|
| 51 | 14.4% | ||
| 52 | 14.2% | ||
| 53 | 13.4% | ||
| 54 |  LongCat 2.0 LongCat | 13.2% | |
| 55 |  gpt-oss-120b (High) OpenAI | 12.8% | |
| 56 | 12.0% | ||
| 57 | 11.8% | ||
| 58 |  K-EXAONE 2.0 LG AI Research | 11.5% | |
| 59 | 10.3% | ||
| 60 |  Qwen3.6 35B A3B (Reasoning) Alibaba | 9.3% | |
| 61 | 8.9% | ||
| 62 |  Nemotron 3.5 Lightning NVIDIA | 8.9% | |
| 63 | 7.6% | ||
| 64 |  LFM2.5-2.6B Liquid AI | 7.2% | |
| 65 | gpt-oss-20b (High) OpenAI | 7.0% | |
| 66 | Qwen3.5 9B (Reasoning) Alibaba | 7.0% | |
| 67 | Qwen3.5 4B (Reasoning) Alibaba | 6.8% | |
| 68 |  Ministral 3 14B Mistral | 6.6% | |
| 69 |  North Mini Code Cohere | 6.4% | |
| 70 | 6.2% | ||
| 71 | Command A+ Cohere | 6.0% | |
| 72 | 6.0% | ||
| 73 |  Trinity Large Thinking Arcee AI | 5.8% | |
| 74 | 5.6% | ||
| 75 |  HyperNova 60B 2605 (High, Based on gpt-oss-120b) Multiverse Computing | 5.6% | |
| 76 | Magistral Medium 1.2 Mistral | 5.6% | |
| 77 | 5.4% | ||
| 78 | Qwen3 Coder Next Alibaba | 5.4% | |
| 79 |  K2 Horizon 0.9B Institute of Foundation Models | 4.9% | |
| 80 | Mistral Small 4 (Reasoning) Mistral | 4.9% | |
| 81 | Ministral 3 3B Mistral | 4.7% | |
| 82 |  KAT Coder Pro V2 KwaiKAT | 4.5% | |
| 83 | Magistral Small 1.2 Mistral | 4.5% | |
| 84 | Qwen3.5 4B (Non-reasoning) Alibaba | 4.3% | |
| 85 |  Celeris-1 Celeris | 3.9% | |
| 86 |  Llama 4 Maverick Meta | 3.7% | |
| 87 | Ministral 3 8B Mistral | 3.7% | |
| 88 | Llama 4 Scout Meta | 3.3% | |
| 89 | gpt-oss-120b (Low) OpenAI | 2.9% |
