𝜏³-Banking
Agents · 2026-03-04
Τ³-Banking is a fintech customer-support benchmark in which agents retrieve policies from a large knowledge base and execute multi-step banking workflows with tools. It measures end-to-end task resolution; higher scores mean more support cases completed.
Top models (higher is better)
| Model | Score |
|---|---|
| Kimi K3 | 34.0 |
| GPT-5.6 Sol | 33.0 |
| Claude Opus 5 | 32.8 |
| Grok 4.5 | 32.6 |
| GPT-5.6 Terra | 31.8 |
| GPT-5.5 | 31.3 |
| DeepSeek-V4-Flash-0731 | 31.1 |
| Sonnet 4.6 | 30.5 |
| GPT-5.4 | 30.3 |
| Opus 4.7 | 28.9 |
| Motif-3-Beta | 28.9 |
| Sonnet 5 | 28.2 |
| JT-4.1 Flash 236B A21B | 28.0 |
| Opus 4.8 | 27.6 |
| GPT-5.6 Luna | 27.2 |
| Claude Fable 5 | 26.8 |
| GLM-5.2 | 26.8 |
| DeepSeek-V4-Pro | 25.8 |
| Gemini 3.5 Flash | 25.4 |
| Muse Spark 1.1 | 25.2 |
| Gemini 3.6 Flash | 24.5 |
| DeepSeek-V4-Flash | 22.9 |
| GPT-5.4 Mini | 21.4 |
| GPT-5.4 Nano | 21.0 |
| Hy3 | 20.8 |
| Kimi K2.6 | 20.6 |
| GPT-5 | 19.6 |
| Muse Spark | 19.6 |
| Sonnet 4.5 | 19.0 |
| DeepSeek-V3.2 | 18.8 |
| Kimi K2.7 Code | 18.1 |
| Qwen3.7-Plus | 17.9 |
| Gemini 3 Flash Preview | 17.5 |
| Nex-N2-Pro | 16.9 |
| Gemini 3.1 Pro Preview | 16.5 |
| Gemini 3.5 Flash-Lite | 16.5 |
| Qwen3.6 Plus (2026-04-02) | 16.5 |
| DeepSeek-V3.1-Terminus | 15.9 |
| Qwen3.6 27B | 15.3 |
| Gemma 4 31B IT | 15.1 |