WildToolBench - Clarification Tasks: leaderboard
Metric: Accuracy (%) on tasks where the correct response is to ask the user for missing information, on WildToolBench (256 multi-turn scenarios, four user tasks each, 1,024 tasks, built from real user-log patterns with expert-annotated tool calls), each model through its native function-call format with default decoding; higher is better. Source: arxiv.org. Saturation forecast: Around 2030. 57 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.0 Flash (Thinking) | 52.34 |
| 2 | O1 | 48.05 |
| 3 | Gemini 2.5 Pro | 46.88 |
| 4 | O3 | 44.92 |
| 5 | GLM-4.5 | 44.53 |
| 6 | DeepSeek R1 | 43.75 |
| 7 | Claude Sonnet 4 | 41.8 |
| 8 | Claude Opus 4.1 | 41.8 |
| 9 | Claude 3.7 Sonnet | 41.41 |
| 10 | Qwen 3 30B A3B (Non-reasoning) | 41.41 |
| 11 | Kimi K2 | 39.84 |
| 12 | Qwen 3 32B (Non-reasoning) | 39.84 |
| 13 | Qwen 3 14B (Non-reasoning) | 39.84 |
| 14 | Qwen 3 8B (Thinking) | 39.84 |
| 15 | GPT-4o | 39.45 |
Interactive version: theaggregate.ai/benchmark?slug=wildtoolbench-clarification-tasks · How It Works · Data refreshed daily, snapshot 2026-10-07.