T1-Bench - Parameter F1: leaderboard

Metric: Parameter F1 (%) of each ground-truth tool call against the best-matching generated call of the same tool; 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

#ModelScore
1Gemma 4 31B (IT)91.06
2Claude Opus 4.689.67
3Gemma 4 26B A4B (IT)89.58
4Claude Haiku 4.589.14
5Claude Sonnet 4.589.1
6gemma-4-E4B-it88.87
7GPT-OSS-120B (High)88.66
8Gemma 4 E2B (IT)88.35
9GPT-5.4 (Low)83.99
10GPT-OSS-20B (High)68.72

Interactive version: theaggregate.ai/benchmark?slug=t1-bench-parameter-f1 · How It Works · Data refreshed daily, snapshot 2026-09-29.