Dharma Field Extractor turns any document into schema-shaped JSON: a 2B-parameter model self-hosted on one L40S, 6 to 147 times cheaper than frontier APIs
Dharma-AI released the Dharma Field Extractor, a 2-billion-parameter model that takes a document image and returns structured JSON, with one checkpoint covering both open extraction and schema-directed requests. On a benchmark of 546 documents across three request scenarios, 1,638 cases in total, it scored 0.834 field-level F1, above the best frontier configuration tested, Claude Opus 5 at 0.816, and the lowest, GPT-5.6 Luna at 0.741; self-hosted on a single L40S, each extraction cost roughly 6 to 147 times less than those APIs.