- New Textract model adds support for 40+ languages and improves long‑word and handwriting accuracy.
- Multilingual table extraction is more robust for merged cells, skewed/rotated, and partial tables.
- No API or response changes needed; the updated model is applied automatically in all regions.