WildHandBench - Formulas: leaderboard
The formula part of WildHandBench (2026): recognition of real-world handwritten formulas, scored by CDM, character detection matching on the rendered formula, in percent; higher is better.
Metric: CDM (%; character detection matching of the rendered formula over the formula documents of 500 real handwritten documents (free text, tables and formulas) in four languages and nine real-world scenarios such as notes, forms, whiteboards and exam sheets). Source: arxiv.org. Saturation forecast: Around February 2027. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 79.42 | #94 |
| 2 | Gemini 3.5 Flash | 78.22 | #101 |
| 3 | Claude Opus 4.8 | 74.11 | #56 |
| 4 | Qwen 3.5 Plus | 73.77 | #232 |
| 5 | Kimi K3 | 70.5 | #54 |
| 6 | Qwen 3 VL 235B A22B (Thinking) | 69.06 | #451 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=wildhandbench-formulas · How It Works · Data refreshed daily, snapshot 2026-11-13.