ToolBench - VirtualHome: leaderboard
ToolBench subtask measuring action generation for VirtualHome household-simulation goals.
Metric: Task Score. Source: huggingface.co. Saturation forecast: Estimated already saturated. 44 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | text-davinci-003 | 25.1 | #1146 |
| 2 | Llama 2 70B | 24.74 | #1201 |
| 3 | CodeLlama-34B-hf | 24.65 | #1515 |
| 4 | CodeLlama-34B-Instruct-hf | 24.34 | #1381 |
| 5 | GPT-4 | 23.5 | #438 |
| 6 | CodeLlama-13B-Instruct-hf | 22.56 | #1532 |
| 7 | CodeLlama-7B-hf | 21.97 | #1457 |
| 8 | CodeLlama-7B-Instruct-hf | 21.95 | #1545 |
| 9 | CodeLlama-13B-Python-hf | 21.94 | #1623 |
| 10 | LLaMA-30B | 21.7 | #1423 |
| 11 | Llama 2 13B | 21.67 | #1436 |
| 12 | Llama 2 7B | 21.49 | #1520 |
| 13 | CodeLlama-7B-Python-hf | 21.33 | #1608 |
| 14 | CodeLlama-34B-Python-hf | 21.24 | #1552 |
| 15 | LLaMA-65B | 20.3 | #1303 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=toolbench-virtualhome · How It Works · Data refreshed daily, snapshot 2026-10-20.