Dharma-AI releases 2B document field extraction model
The story so far
Dharma-AI has released the Dharma Field Extractor, a 2B parameter model: input document images and optional schema, output typed structured JSON, supports open extraction and schema-directed extraction, fields not appearing in the document return null. The same checkpoint covers all document types, no need for type-specific fine-tuning. Dharma-AI claims the model outperforms all front-line LLMs it has tested in testing, at only a small fraction of the cost. The model runs self-hosted, not charged per token; its stated cost metric is 2.24 US dollars per hour of GPU time for a single L40S, and notes this is not a sales price, actual self-hosted costs depend on hardware, usage, and deployment latency.
AI-written from the reports · Updated 14 days ago
Coverage timeline
Follow the coverage to see every side of the story.
- Hugging Face BlogFeaturedDharma Field Extractor: Any Document In, Structured Fields in JSON Out
Dharma-AI releases Dharma Field Extractor, a 2B parameter model, input document image and optional schema, output typed structured JSON, supports open extraction and schema-directed extraction, missing fields return null.