Excel Modeling Benchmark
Professional Work · 2026-07-01
Excel Modeling Benchmark evaluates agents on Excel-based financial modeling tasks, including template-based and scratch financial model construction scored against expert-authored reference workbooks and rubrics.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5.1 | 76.7 |
| Claude Opus 5.5 | 75.9 |
| Claude Sonnet 5.5 | 75.7 |
| Gemini 4 Argon | 75.2 |
| Claude Fable 5 | 73.7 |
| Claude Opus 5 | 73.6 |
| GPT-5.6 Sol | 72.3 |
| Gemini 3.8 Flash | 72.2 |
| Qwen3.8 Max (0902) | 72.0 |
| Muse Wonder Full | 71.9 |
| GPT-6 Astra | 71.7 |
| GPT-6 Sol | 71.5 |
| Gemini 3.7 Flash | 71.3 |
| GPT-6.1 Sol | 70.8 |
| Opus 4.8 | 69.4 |
| GPT-6 Luna | 68.5 |
| Muse Spark 1.3 | 67.4 |
| GPT-5.6 Luna | 67.1 |
| Muse Wonder | 67.1 |
| Grok 4.7 | 67.0 |
| Kimi K3 | 66.7 |
| Sonnet 5 | 66.3 |
| GPT-5.6 Terra | 66.2 |
| MiMo-V2.6-Flash | 65.5 |
| Gemini 3.6 Flash | 65.4 |
| GPT-5.5 | 64.5 |
| Opus 4.7 | 63.7 |
| Gemini 3.5 Flash | 63.5 |
| Ember-1 | 63.4 |
| MiMo-V2.6-Pro | 62.9 |
| Grok 4.6 | 62.7 |
| DeepSeek V4.1 Flash 1017 | 61.8 |
| GLM-5.2 | 61.5 |
| Sonnet 4.6 | 60.2 |
| Qwen3.8-Max | 60.1 |
| Qwen3.8 27B | 59.7 |
| Claude Haiku 5.5 | 59.0 |
| Hy4 Preview | 58.1 |
| Kimi K2.6 | 57.9 |
| DeepSeek V4.1 Flash | 57.2 |