Studio Dots AI

license

MIT(OSI

weights

open(published on Hugging Face and downloadable without a gate)

data

closed(the training corpus is neither published nor described)

code

partial(inference code and a loader are published

MIT covers both the code and the weights, which download freely from Hugging Face. But the repository publishes inference code only, not the training pipeline, and the training corpus is neither published nor described, so this reads as an open-weights release rather than a fully open model.

• https://github.com/studio-dots-ai/dots.ocr recorded 2026-09-16

  Repository page for studio-dots-ai/dots.ocr: MIT, public and unarchived, 9,117 stars, carrying inference code with the weights published on Hugging Face. Establishes the licence, the published code and the absence of a training pipeline.

Adoption is measured as Hugging Face downloads of the dots.ocr checkpoint over the trailing month, the only usage-volume channel this record declares.

• https://huggingface.co/api/models/dots-studio/dots.ocr recorded 2026-09-16

  Hugging Face model API record for dots-studio/dots.ocr: 668,858 downloads in the trailing 30 days, license mit. Establishes the usage-volume channel and its figure.

dots.ocr's model card documents layout detection, reading order and table recognition, all emitted as JSON. Its capability matches Chandra's, the other single-model parser here.

• https://github.com/studio-dots-ai/dots.ocr recorded 2026-09-16

  Repository page and README describing unified layout detection and content recognition in one vision-language model with prompt-based task switching, and table recognition emitted as JSON.