WildToolBench - Single-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 the tasks solvable with one tool call; higher is better.

Metric: Accuracy (%) on tasks solvable with one tool call, 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 May 2028. 50 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1xLAM-2-70B64.45
2toolace2-8B62.11
3xLAM-2-8B62.11
4O361.72#242
5watt-tool-8B61.72
6xLAM-2-32B60.94
7GPT-4o60.16#754
8Claude Sonnet 460.16#414
9Qwen 3 8BNon-reasoning60.16#1890
10Qwen 3 4BReasoning60.16#2379
11Doubao-1.5-Thinking-Pro60.16#265
12Grok 459.38#321
13Qwen 2.5 72B Instruct58.98#1124
14DeepSeek V358.98#736
15Doubao-1.5-Pro58.59#613

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=wildtoolbench-single-tool-tasks · How It Works · Data refreshed daily, snapshot 2026-11-22.