MortgageTax: leaderboard

Evaluating reading and understanding tax certificates as images.

Metric: Accuracy (%). Source: vals.ai. 98 models tracked.

Top models

#ModelScoreOverall rank
1Claude Opus 572.06#12
2Claude Fable 5.170.79#9
3Claude Opus 4.770.27#49
4Claude Sonnet 570.03#57
5Claude Opus 4.869.91#37
6Gemini 3.1 Pro (Preview)69.4#58
7Gemini 3 Pro (Preview)69.08#66
8Gemini 2.5 Pro68.92#152
9Claude Fable 568.92#16
10GPT-5.568.76#31
11Gemini 3 Flash (Preview)68.72#82
12Claude 3.7 Sonnet (20250219)68.68#201
13Gemini 3.5 Flash Lite68.68#156
14Claude Opus 4.5 (20251101)68.68#74
15Claude Opus 4.6 (Thinking)68.52#63 (Claude Opus 4.6)

No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.

Interactive version: theaggregate.ai/benchmark?slug=mortgagetax · How It Works · Data refreshed daily, snapshot 2026-10-20.