ToolBench - VirtualHome — leaderboard

Metric: Task Score. Source: huggingface.co. 44 models tracked.

Top models

#ModelScore
1text-davinci-00325.1
2Llama 2 70B24.74
3CodeLlama-34B-hf24.65
4CodeLlama-34B-Instruct-hf24.34
5GPT-423.5
6CodeLlama-13B-Instruct-hf22.56
7CodeLlama-13B-hf22.28
8CodeLlama-7B-hf21.97
9CodeLlama-7B-Instruct-hf21.95
10CodeLlama-13B-Python-hf21.94
11LLaMA-30B21.7
12Llama 2 13B21.67
13Llama 2 7B21.49
14CodeLlama-7B-Python-hf21.33
15CodeLlama-34B-Python-hf21.24

Interactive version: theaggregate.ai/benchmark?slug=toolbench-virtualhome · How the rankings work · Data refreshed daily, snapshot 2026-07-22.