Excel-Bench (Humanlaya): leaderboard
Humanlaya spreadsheet-manipulation benchmark: models solve real Excel tasks; average score %, with OpenAI effort tiers reported as separate rows.
Metric: Average Score (%). Source: humanlaya.com. Saturation forecast: Around February 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 (xHigh) | 72.92 | #83 (GPT-5.4) |
| 2 | GPT-5.4 (High) | 72.74 | #83 (GPT-5.4) |
| 3 | Claude Opus 4.7 | 66.01 | #51 |
| 4 | Claude Opus 4.6 | 65.97 | #67 |
| 5 | Kimi K2.5 | 44.27 | #152 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=excel-bench-humanlaya · How It Works · Data refreshed daily, snapshot 2026-11-05.