DeepSeek-OCR
No written verdict yet. Here is what commenters said.
6recommend
2warn against
Nov 1, 2025first mention
Oct 25, 2026latest mention
Why people recommend it
“check out PEXEL and DeepSeek-OCR.”
“Deepseek was the first to really jump on this”
Show 3 more
“What makes the DeepSeek-OCR and related results exciting to some researchers”
“DeepSeek-OCR proved 10x+ compression from visual embedding of text”
“I love that the entire goal is to push Deepseek OCR further.”
Caveats and dissent
“PaddleOCR, dots.ocr and DeepSeek OCR performed significantly worse”
“Deepseek OCR is no longer state of the art.”
Often compared with
Threads where it came up
- 60% Fable cost cut by converting code to images and having the model OCR it 2 mentions
- Dessa: Conversation Model trained only on speech 1 mention
- OCR It – pull text out of un-copyable documents for your LLM 1 mention
- How Compaction Works in Pi 1 mention
- Unlimited OCR: One-Shot Long-Horizon Parsing 1 mention
- Rolling your own serverless OCR in 40 lines of code 1 mention
Also written as DeepSeek OCR. Data: deepseek-ocr.json. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.
All comments 9
“check out PEXEL and DeepSeek-OCR.”
“PaddleOCR, dots.ocr and DeepSeek OCR performed significantly worse”
“Deepseek was the first to really jump on this”
“What makes the DeepSeek-OCR and related results exciting to some researchers”
“DeepSeek-OCR proved 10x+ compression from visual embedding of text”
“I love that the entire goal is to push Deepseek OCR further.”
“Deepseek OCR is no longer state of the art.”
“Built on DeepSeek-OCR”
“images can be profoundly more token-efficient encodings of information than even CSVs”