When2Tool (Single-Hop): leaderboard

Metric: Accuracy (%; exact-form answers in a boxed final response, checked by a deterministic evaluator; simulated offline tool environments at easy, medium and hard difficulty; 2,250 single-hop test tasks in 15 environments; tools offered under the neutral default prompt; mean of three seeds). Source: arxiv.org. Saturation forecast: Estimated already saturated. 6 models tracked.

Top models

#ModelScore
1Qwen 3 32B94.1
2Qwen 3 14B93.7
3Qwen 3 4B 2507 Instruct89.2
4Qwen 3 1.7B88.2
5Llama 3.3 70B Instruct83.1
6Llama 3.1 8B79.5

Interactive version: theaggregate.ai/benchmark?slug=when2tool-single-hop · How It Works · Data refreshed daily, snapshot 2026-09-26.