Dharma Field Extractor turns documents into structured JSON

Dharma-AI has introduced Dharma Field Extractor, a 2-billion-parameter model designed to pull requested information from document images and return it as structured JSON. It can operate without a predefined schema, identifying fields from labels, values and table layouts, or follow an exact schema with specified names and data types. Missing fields are returned as null, while tables are represented row by row, including nested structures.

The system is based on DharmaOCR-LITE v1.1 and uses a LoRA adapter trained on 43,384 examples. The training material covered English documents, internally sourced Brazilian documents in Portuguese, and a smaller multilingual collection. Examples included receipts, invoices, forms, bank statements, identity documents and synthetic documents.

The reported benchmark contained 546 documents and 1,638 cases across three extraction settings. Dharma Field Extractor scored 0.834, compared with 0.816 for Claude Opus 5 and 0.741 for GPT-5.6 Luna. The company said its measured running cost was approximately six to 147 times lower than the tested APIs. Its figure was based on self-hosting one L40S GPU at $2.24 per hour, rather than an API sales price, so actual costs depend on hardware, volume and latency requirements.