Maternity Case Record Digitisation (Synthetic Forms): leaderboard
A 2026 benchmark of multimodal LLMs for digitising a handwritten South African maternity case record: each form image is extracted into a JSON template and compared field by field with reference JSON files. This board is field-level exact-match accuracy in percent, empty fields included, median over four synthetic forms in the same layout; higher is better.
Metric: Field-level accuracy (from 0-1, times 100; median over forms): share of form fields, empty fields included, whose normalised extraction exactly matches the ground truth, on four synthetic Maternity Case Records filled with simulated data in the same layout as the real registry form, one form image per call with the optimised extraction prompt and JSON template, temperature 0; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 16 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Gemini 3.1 Flash Lite | 86 | #377 | |
| 2 | Gemini 3.1 Pro (Preview) | 84 | #102 | |
| 3 | GPT-5.4 | 80 | #138 | |
| 4 | Gemini 2.5 Pro | 79 | #280 | |
| 5 | Claude Sonnet 4.6 | 77 | #179 | |
| 6 | GPT-5.2 | 76 | #190 | |
| 7 | GPT-5.2 Pro | 76 | #139 | |
| 8 | Gemini 2.5 Flash | 75 | #466 | |
| 9 | Claude Opus 4.6 | 73 | #114 | |
| 10 | GPT-4.1 | 67 | #506 | |
| 11 | Grok 4.1 Fast | Reasoning | 67 | #431 |
| 12 | Claude Haiku 4.5 | 62 | #587 | |
| 13 | GPT-5 Mini | 60 | #373 | |
| 14 | Qwen 2.5 VL 3B | 48 | #2664 | |
| 15 | GPT-5 Nano | 46 | #912 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=maternity-case-record-digitisation-synthetic-forms · How It Works · Data refreshed daily, snapshot 2026-11-22.