SpreadsheetBench v2 - Template Accuracy: leaderboard
Metric: Task accuracy (%). Source: spreadsheetbench.github.io. Saturation forecast: Around January 2027. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.6 | 52.58 |
| 2 | GPT-5.2 | 35.05 |
| 3 | Gemini 3.1 Pro (Preview) | 28.87 |
| 4 | DeepSeek V3.2 | 25.77 |
| 5 | Kimi K2.5 | 18.56 |
| 6 | Qwen 3.5 397B A17B | 17.53 |
| 7 | GLM-5 | 17.53 |
| 8 | MiniMax-M2.5 | 7.22 |
Interactive version: theaggregate.ai/benchmark?slug=spreadsheetbench-v2-template-accuracy · How It Works · Data refreshed daily, snapshot 2026-10-02.