OCR models / rednote (dots)

dots.ocr

Name: dots.ocr
Author: rednote (dots)

rednote (dots) · released 2025-07-30 · data checked 2026-11-06

dots.ocr is a 3B-parameter OCR and document-parsing model from rednote (dots) (CN), released 2025-07-30 under the MIT license. It scores 90.77 on OmniDocBench v1.6 (#25 of 33) and 79.1 on olmOCR-Bench (#15 of 24). It needs about 7.1 GB of memory at 16-bit.

Widely used; succeeded by dots.mocr.

OmniDocBench v1.6

90.77

#25 of 33 models

olmOCR-Bench

79.1

#15 of 24 models

Parameters

OCR specialist

Memory to run

~7.1 GB

16-bit weights plus about 1 GB

Benchmarks

OmniDocBench v1.6 Leaderboard90.77

Best: OXR 1.0 97.03

olmOCR-Bench Leaderboard79.1

Best: OXR 1.0 87.8

The tick marks the best score on each benchmark. See how to read the ranking.

Specifications

Lab: rednote (dots) (CN)
Type: OCR / document parsing specialist
Total parameters: 3B
License: MIT
Memory to run: ~7.1 GB at 16-bit
Released: 2025-07-30

Weights and usage on Hugging Face ↗

Closest alternatives

jina-ocr-v1 (3.4B, 91.14 on OmniDocBench, CC-BY-NC-4.0)
DeepSeek-OCR 2 (3.4B, 90.25 on OmniDocBench, Apache-2.0)
HunyuanOCR 1.0 (1B, 89.95 on OmniDocBench, Tencent Hunyuan Community License)
MonkeyOCR v2 (B) (880M, 91.80 on OmniDocBench, Apache-2.0)

Sources

Something wrong here? Report it.