# Hugging Face — Dharma Field Extractor: Any Document In, Structured Fields in JSON Out

- Company: Hugging Face (huggingface.co)
- Announced: 2026-11-04T17:50:10+00:00
- Category: new-model
- Coverage: 1 outlet
- Announcement: yes
- Group: models
- Source: https://huggingface.co/blog/Dharma-AI/dharma-field-extractor
- Record: https://forck.live/items/31680-dharma-field-extractor-any-document-in-structured-fields-in-json-out
- Subject: Platform
- Models affected: Dharma Field Extractor, DharmaOCR-LITE, DharmaOCR-LITE v1.1, Claude Opus 5, GPT-5.6 Luna, GPT-5.6 Sol

Dharma AI released the Dharma Field Extractor, a 2-billion-parameter model that pulls requested fields out of any document type and returns them as structured JSON. The company says the model outperformed every frontier LLM it tested at a fraction of the cost, addressing a task where OCR returns a full page of text with no indication of which string is the exam type, the date or the total. The model accepts either an open request, where it infers structured fields from layout and context, or a schema that fixes the exact keys and types, returning null for fields absent from the document. It is built on DharmaOCR-LITE v1.1, a new 2-billion-parameter checkpoint, with a LoRA adapter trained on 43,384 examples spanning English multi-domain, Brazilian Portuguese and multilingual data. The benchmark covers 546 documents and 1,638 cases across three request scenarios, scored by field-level F1 adjusted for value similarity and valid JSON output. Against four general-purpose frontier configurations, Dharma says the quality gap runs from 1.8 points for Claude Opus 5 to 9.3 points for GPT-5.6 Luna, while cost runs from roughly 6 times lower to roughly 147 times lower.

## Evidence

Verbatim from https://huggingface.co/blog/Dharma-AI/dharma-field-extractor:

> A 2-billion-parameter model that extracts exactly the fields you ask for, from any document type, outperforming every frontier LLM we tested, at a fraction of the cost.
> 
> In April, we released DharmaOCR-LITE , a model specialized for OCR across a wide range of documents, from contracts and court filings to invoices, IDs and forms. The paper behind it made a narrow, well-measured case for domain specialization: a smaller model trained closer to its task beating general-purpose frontier APIs on that task.
> 
> OCR gives you the text on a page, faithful to the original formatting. That's exactly what you want when the job is the whole document, like a contract you'll read end to end or a manuscript you're digitizing.
> 
> However, there's a different use case: you don't want the page, you want specific fields out of it, in a shape a system downstream can consume. Every line item on an invoice, as a list. The patient name, exam type and date from a medical requisition, ready to schedule. The fields off an ID, mapped to the columns of an employee record.

## Around this story

Outlets this record can name and link:

- Extrapolator — Dharma Field Extractor: Any Document In, Structured Fields in JSON Out: https://extrapolator.ai/2026/11/04/dharma-field-extractor-any-document-in-structured-fields-in-json-out

---

Record: https://forck.live/items/31680-dharma-field-extractor-any-document-in-structured-fields-in-json-out
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md