Benchmark
Tau3 Banking
τ³-Bench banking domain evaluates agentic models on multi-turn, tool-using customer-support scenarios in a simulated retail banking environment.
- Modality
- text
- Categories
- reasoning, agents, tool_calling
- Openness
- unknown
- Source
- llm_stats
- Reported scores
- 7
Reported scores
llm_stats
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| Grok 4.5 | xAI | 0.33 | 2026-07-16 | evidence |
| Inkling-Small | Thinking Machines Lab | 0.155 | 2026-07-30 | evidence |
| LFM2.5-2.6B | Liquid AI | 0.0567 | 2026-08-04 | evidence |
| Mistral Medium 3.5 | Mistral | 0.134 | 2026-04-29 | evidence |
| Muse Glimmer-30B | Meta | 0.235 | 2026-08-10 | evidence |
| Nemotron 3.5 Lightning (30B A3B) | NVIDIA | 0.0928 | 2026-08-11 | evidence |
| Solar Pro 4 | Upstage | 0.23 | 2026-08-06 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.