WildToolBench - Task Accuracy: leaderboard
Metric: Task accuracy (%) over all 1,024 tasks, 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) | 61.04 |
| 2 | O1 | 58.79 |
| 3 | O3 | 56.93 |
| 4 | Claude Sonnet 4 | 56.54 |
| 5 | Gemini 2.5 Pro | 56.25 |
| 6 | GLM-4.5 | 56.05 |
| 7 | DeepSeek R1 | 55.27 |
| 8 | Claude Opus 4.1 | 54.98 |
| 9 | GPT-4o | 54.88 |
| 10 | GPT-4.1 | 54.3 |
| 11 | Qwen 3 8B (Thinking) | 54.3 |
| 12 | Qwen 3 14B (Non-reasoning) | 54.1 |
| 13 | Qwen 3 4B (Reasoning) | 53.91 |
| 14 | Kimi K2 | 53.71 |
| 15 | Qwen 3 14B (Reasoning) | 53.22 |
Interactive version: theaggregate.ai/benchmark?slug=wildtoolbench-task-accuracy · How It Works · Data refreshed daily, snapshot 2026-10-07.