WildToolBench - Multi-Tool Tasks: leaderboard

WildToolBench (2026) evaluates LLM tool use under realistic user behavior: 256 multi-turn scenarios over about 1,600 public APIs mix compositional tool orchestration, intentions hidden in earlier turns and switches between tool, chat and clarification requests. This board is accuracy in percent on compositional tasks that need several sequential or parallel tool calls; higher is better.

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 November 2027. 50 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-4.144.14#505
2Claude Sonnet 443.75#410
3GPT-4o41.8#751
4Grok 441.41#319
5DeepSeek R141.02#482
6DeepSeek V3.140.63#511
7GLM-4.540.63#542
8Seed-1.640.23#469
9Gemini 2.0 Flash (Thinking)40.23
10Claude Opus 4.139.84#278
11O339.45#237
12Claude 3.7 Sonnet39.06#470
13O139.06#403
14DeepSeek V338.67#729
15Qwen 2.5 32B Instruct38.67#1247

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=wildtoolbench-multi-tool-tasks · How It Works · Data refreshed daily, snapshot 2026-11-21.