ToolBench - Google Sheets: leaderboard
ToolBench subtask measuring spreadsheet tool-use success on Google Sheets-style actions.
Metric: Task Score. Source: huggingface.co. Saturation forecast: Estimated already saturated. 44 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | CodeLlama-34B-hf | 64.29 | #1515 |
| 2 | GPT-4 | 62.9 | #438 |
| 3 | text-davinci-003 | 62.9 | #1146 |
| 4 | CodeLlama-34B-Instruct-hf | 61.11 | #1381 |
| 5 | Llama 2 70B | 58.57 | #1201 |
| 6 | CodeLlama-34B-Python-hf | 55.87 | #1552 |
| 7 | GPT-3.5 Turbo | 51.4 | #865 |
| 8 | CodeLlama-13B-Python-hf | 50.79 | #1623 |
| 9 | CodeLlama-7B-Python-hf | 49.13 | #1608 |
| 10 | CodeLlama-13B-Instruct-hf | 48.97 | #1532 |
| 11 | starcoder | 48 | #1577 |
| 12 | CodeLlama-7B-Instruct-hf | 41.27 | #1545 |
| 13 | CodeLlama-7B-hf | 38.08 | #1457 |
| 14 | LLaMA-30B | 37.1 | #1423 |
| 15 | LLaMA-65B | 32.9 | #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-google-sheets · How It Works · Data refreshed daily, snapshot 2026-10-20.