T1-Bench - Tool Output Match: leaderboard

Metric: Exact-match rate (%) of ground-truth tool outputs reproduced by some tool call in the conversation; 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: Around April 2027. 12 models tracked.

Top models

#ModelScore
1Gemma 4 31B (IT)82.85
2Gemma 4 26B A4B (IT)79.88
3gemma-4-E4B-it79.88
4Claude Sonnet 4.579.28
5Claude Opus 4.679
6Claude Haiku 4.578.95
7GPT-OSS-120B (High)78.19
8Gemma 4 E2B (IT)76.39
9GPT-5.4 (Low)68.89
10GPT-OSS-20B (High)57.49

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