WildToolBench - Multi-Tool Tasks: leaderboard
Metric: Accuracy (%) on compositional tasks needing multi-step sequential or parallel tool calls, 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 March 2028. 57 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4.1 | 44.14 |
| 2 | Claude Sonnet 4 | 43.75 |
| 3 | GPT-4o | 41.8 |
| 4 | Grok 4 | 41.41 |
| 5 | DeepSeek R1 | 41.02 |
| 6 | DeepSeek V3.1 | 40.63 |
| 7 | GLM-4.5 | 40.63 |
| 8 | Seed-1.6 | 40.23 |
| 9 | Gemini 2.0 Flash (Thinking) | 40.23 |
| 10 | Claude Opus 4.1 | 39.84 |
| 11 | O3 | 39.45 |
| 12 | Claude 3.7 Sonnet | 39.06 |
| 13 | O1 | 39.06 |
| 14 | DeepSeek V3 | 38.67 |
| 15 | Qwen 2.5 32B Instruct | 38.67 |
Interactive version: theaggregate.ai/benchmark?slug=wildtoolbench-multi-tool-tasks · How It Works · Data refreshed daily, snapshot 2026-10-07.