DeepSeek-OCR

deepseek-ai/DeepSeek-OCR

Image Text to Text3.3B parameterstransformersLicense: mit
Tracked by OpenModelStats since September 5, 2026View on Hugging Face ↗
Downloads 30D
2.1M
Total downloads
36.3M
Gained 7D
+497.2K(increase)
1.4%(up) vs baseline
Gained 30D
—
Likes
3,391
+8 in 7D
Rank · tracked models
#185
2(down 2 places)this week
Parameters
3.3B
Last source update
Nov 4, 2025
Momentum
89.3
  • DeepSeek-OCR gained 497.2K downloads during the last seven tracked days.
  • It moved from #26 to #27 among tracked Image Text to Text models this week.

Downloads over time

Cumulative total downloads observed by OpenModelStats

Likes over time

Rank among tracked models

Lower is better

Overview

DeepSeek-OCR is a image text to text model published by deepseek-ai. OpenModelStats has tracked the model since Sep 5, 2026, most recently observing it Oct 5, 2026.

Published
Oct 17, 2025
Last updated
Nov 4, 2025
Library
transformers
License
mit
Task
Image Text to Text
Parameters
3,336,106,240
Spaces
—
Derivative models
2
Velocity 7D/day
71K
transformerssafetensorsdeepseek_vl_v2feature-extractiondeepseekvision-languageocrcustom_codeimage-text-to-textmultilingualarxiv:2510.18234license:mit

Model family

Based on base-model relationships reported by the source's metadata.

Similar models

Models similar to DeepSeek-OCR
ModelDL 30D
Unlimited-OCRbaidu1.4M
DeepSeek-OCR-8bitmlx-community1,783
DeepSeek-OCRunsloth1,172
Unlimited-OCR-Community-FP8-Dynamicshadowrock-io609
Unlimited-OCR-mxfp8mlx-community436
Unlimited-OCR-8bitmlx-community409