MortgageTax - Numerical Extraction
Professional Work · 2025-03-05
The Numerical Extraction split of MortgageTax. This child benchmark separates a source-reported subtask or subtrack from the parent aggregate so scores at different grains do not share one benchmark_slug.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5 | 81.4 |
| Opus 4.8 | 79.7 |
| GPT-5.4 Mini | 79.4 |
| Kimi K3 | 78.9 |
| Opus 4.7 | 78.7 |
| Sonnet 5 | 78.7 |
| Qwen3.6 Plus (2026-04-02) | 78.1 |
| Claude 3.7 Sonnet | 77.7 |
| Gemini 2.5 Pro | 77.6 |
| Gemini 3.5 Flash | 77.5 |
| Opus 4.5 | 77.4 |
| MiniMax M3 | 77.3 |
| Gemini 3.1 Flash-Lite Preview | 77.0 |
| Opus 4.6 | 76.9 |
| Claude 3.5 Sonnet (Oct 2024) | 76.9 |
| Gemini 3.6 Flash | 76.8 |
| GPT-5 | 76.8 |
| Claude Fable 5 | 76.7 |
| GPT-5.2 | 76.5 |
| Gemini 3.1 Pro Preview | 76.4 |
| GPT-4.1 | 76.2 |
| GPT-5.5 | 76.2 |
| Gemini 3 Pro Preview | 76.0 |
| GPT-5.4 | 76.0 |
| Gemini 3.5 Flash-Lite | 75.8 |
| Gemini 3 Flash Preview | 75.8 |
| Qwen3.6 27B | 75.8 |
| GPT-5.6 Luna | 75.6 |
| GPT-5 Mini | 75.4 |
| GPT-5.1 | 75.2 |
| GPT-5.6 Sol | 75.1 |
| Sonnet 4.6 | 75.0 |
| Kimi K2.5 | 75.0 |
| Qwen3.5-Flash | 75.0 |
| Qwen3.7-Plus | 74.8 |
| Gemini 2.5 Pro Experimental 03-25 | 74.6 |
| Kimi K2.6 | 74.5 |
| Muse Spark 1.1 | 74.5 |
| GPT-5.6 Terra | 74.3 |
| o3 | 73.8 |