Excel Modeling Benchmark Template
Professional Work · 2026-07-01
Template-mode score on Excel Modeling Benchmark, where agents complete a stripped template workbook and are graded against the gold workbook.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 81.4 |
| Claude Sonnet 5.5 | 80.9 |
| Claude Fable 5.1 | 80.8 |
| Gemini 4 Argon | 79.6 |
| Claude Fable 5 | 77.0 |
| Gemini 3.8 Flash | 75.6 |
| Qwen3.8 Max (0902) | 75.4 |
| GPT-5.6 Sol | 75.3 |
| GPT-6.1 Sol | 75.1 |
| GPT-6 Astra | 74.8 |
| Claude Opus 5 | 74.7 |
| Gemini 3.7 Flash | 73.6 |
| GPT-6 Sol | 73.5 |
| Muse Wonder Full | 73.3 |
| Opus 4.8 | 71.1 |
| GPT-6 Luna | 70.4 |
| GPT-5.6 Terra | 68.5 |
| Sonnet 5 | 68.0 |
| Kimi K3 | 66.7 |
| Muse Wonder | 65.7 |
| Grok 4.7 | 65.6 |
| Muse Spark 1.3 | 65.5 |
| MiMo-V2.6-Pro | 64.9 |
| Gemini 3.6 Flash | 64.7 |
| GPT-5.6 Luna | 64.7 |
| MiMo-V2.6-Flash | 64.6 |
| Claude Haiku 5.5 | 64.3 |
| Gemini 3.5 Flash | 62.5 |
| GPT-5.5 | 61.8 |
| Ember-1 | 60.8 |
| Opus 4.7 | 60.4 |
| GLM-5.2 | 55.8 |
| Kimi K2.6 | 55.7 |
| Sonnet 4.6 | 55.2 |
| Qwen3.8-Max | 55.1 |
| Grok 4.6 | 54.7 |
| Qwen3.8 27B | 54.7 |
| DeepSeek V4.1 Flash 1017 | 54.7 |
| MiMo-V2.5 | 52.2 |
| DeepSeek-V4-Flash-0731 | 52.2 |