Event date: 2026-11-04
Dharma-AI introduced Field Extractor on November 4: a two-billion-parameter model for turning document images into typed JSON. It supports both open-ended extraction and a specified schema.
The team says absent fields are returned as null and tables preserve their field structure. Its published comparison uses 546 documents across three request modes. Those are task-specific, developer-reported results, not proof that the model outperforms general models on every document or language.
Why it matters
The appeal is a smaller component for a narrow job. Before connecting it to records or payments, evaluate your own scans, missing fields, table totals and rejected outputs rather than checking JSON syntax alone.
Source: Dharma-AI on Hugging Face · Checked November 6, 2026