T1-Bench - Tool Call F1: leaderboard
Metric: Tool-name F1 (%) of the assistant's tool calls against the ground-truth calls, repeated calls counted individually; T1-Bench: 525 multi-turn customer-service conversations over 25 domain settings (11 single domains and 14 combinations of 2 to 11 domains, 76 tools), a Gemma4-26B-A4B-it user simulator, three trials per conversation at temperature 1.0; averages weighted by conversations per setting; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.6 | 93.34 |
| 2 | Gemma 4 31B (IT) | 92.59 |
| 3 | Claude Haiku 4.5 | 92.41 |
| 4 | Claude Sonnet 4.5 | 92.39 |
| 5 | Gemma 4 26B A4B (IT) | 90.48 |
| 6 | gemma-4-E4B-it | 90.33 |
| 7 | Gemma 4 E2B (IT) | 90.22 |
| 8 | GPT-OSS-120B (High) | 89.95 |
| 9 | GPT-5.4 (Low) | 87.8 |
| 10 | GPT-OSS-20B (High) | 73.51 |
Interactive version: theaggregate.ai/benchmark?slug=t1-bench-tool-call-f1 · How It Works · Data refreshed daily, snapshot 2026-09-29.