UserToolBench: leaderboard

Metric: Exact trajectory accuracy (%; mean of the single-tool, multi-tool and lack-of-information task-type accuracies; a trajectory counts when it matches the profile-conditioned reference: the tool name and decision-relevant arguments, the ordered call sequence for multi-tool tasks, or the correct clarify-or-infer decision when information is missing; 799 profile-hidden evaluation instances from long-horizon interaction trajectories of 10 privacy-sanitized user profiles over 36 tool sets; the model sees the interaction history, the current request and the tool schemas but not the profile). Source: arxiv.org. Saturation forecast: Around 2029. 9 models tracked.

Top models

#ModelScore
1GLM-549.36
2DeepSeek V4 Pro42.49
3Kimi K2.638.65
4Qwen 3.6 Plus37.02
5Gemini 3.5 Flash34.32
6GPT-5.431.98

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