Excel Modeling Benchmark Scratch
Professional Work · 2026-07-01
Scratch-mode score on Excel Modeling Benchmark, where agents build a financial model without a template workbook and are judged against the task rubric.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5.1 | 72.6 |
| Claude Opus 5 | 72.3 |
| Gemini 4 Argon | 70.8 |
| Grok 4.6 | 70.8 |
| Muse Wonder Full | 70.5 |
| Claude Sonnet 5.5 | 70.5 |
| Claude Opus 5.5 | 70.5 |
| Claude Fable 5 | 70.0 |
| GPT-6 Sol | 69.5 |
| GPT-6 Astra | 69.1 |
| GPT-5.6 Sol | 69.1 |
| Muse Wonder | 69.0 |
| Muse Spark 1.3 | 68.8 |
| GPT-5.6 Luna | 68.8 |
| DeepSeek V4.1 Flash 1017 | 68.5 |
| Qwen3.8 Max (0902) | 68.5 |
| Gemini 3.8 Flash | 68.4 |
| Gemini 3.7 Flash | 68.2 |
| Grok 4.7 | 67.8 |
| Hy4 Preview | 67.6 |
| Opus 4.8 | 67.4 |
| Grok 4.5 | 67.3 |
| Opus 4.7 | 67.0 |
| GPT-5.5 | 66.8 |
| Step 5 Preview | 66.8 |
| GLM-5.2 | 66.7 |
| Kimi K3 | 66.5 |
| GLM-5.3 | 66.5 |
| GPT-6 Luna | 66.4 |
| GPT-6.1 Sol | 66.4 |
| MiMo-V2.6-Flash | 66.1 |
| DeepSeek V4 Pro 0813 | 66.1 |
| DeepSeek V4.1 Flash | 66.1 |
| Gemini 3.6 Flash | 65.7 |
| Ember-1 | 65.6 |
| Sonnet 5 | 65.0 |
| Sonnet 4.6 | 64.8 |
| Mistral Large 4 Preview | 64.7 |
| Qwen3.8-Max | 64.5 |
| Qwen3.8 27B | 64.1 |