Skip to content
CAI
Software that uses CAICheck a score

deepseek-ai/DeepSeek-OCR

42.3

Weak · 16 November 2026

2.5k

lines of production code

Python

primary language

5

measurements over time

CAI band scale
CAI trend line
CAI lens gauges

What this system is

This system is a codebase for the DeepSeek-OCR model, providing inference and serving capabilities for document image analysis. It integrates with Hugging Face Transformers and vLLM to handle multi-modal processing, including custom image preprocessing and specialized logits processors. The current scope is limited to the initial release of these core components and associated evaluation scripts.

How it got here

October 2025 — Initial release of DeepSeek-OCR inference and serving components

1 change.

This commit introduces the initial codebase for the DeepSeek-OCR model, providing both a Hugging Face Transformers inference script and a vLLM-based serving implementation. The vLLM integration includes the core model definition, a custom multi-modal processor for image handling, and specific image preprocessing logic (including dynamic cropping and resizing). It also adds a no-repeat n-gram logits processor to improve generation quality and several runner scripts for batch evaluation, single image inference, and PDF processing.

Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

How this codebase got here

Score

  • CAI 41 → 42 (+1.2)
  • Rubric changed (rubric-2026.11.2 → rubric-2026.11.4) — scores are not directly comparable.

Lenses

  • Code Health 73 → 72 (-1.5)
  • Architecture 99 → 99 (+0.0)
  • Maturity 47 → 52 (+5.0)
  • Readiness 15 → 15 (+0.0)
  • Security 71 → 71 (+0.0)
  • Performance 100 → 100 (+0.0)

Resolved (4)

  • Duplicated block (6 lines × 2) (DeepSeek-OCR-master/DeepSeek-OCR-vllm/run_dpsk_ocr_image.py)
  • Duplicated block (8 lines × 2) (DeepSeek-OCR-master/DeepSeek-OCR-vllm/run_dpsk_ocr_image.py)
  • Near-duplicate member pair (57 shared lines) (DeepSeek-OCR-master/DeepSeek-OCR-vllm/run_dpsk_ocr_image.py)
  • No tests found

New (1)

  • Edited copy of a member (63 corresponding lines) (DeepSeek-OCR-master/DeepSeek-OCR-vllm/run_dpsk_ocr_image.py)

Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

Survey your own repository

deepseek-ai/DeepSeek-OCR was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.

About this page

  • The score is its most recent published measurement, taken on 16 November 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
  • Measured at commit 09eaf526153e7a01ed16c9dea8c96282aaea29c0 — the exact code this score is about.
  • Scored under rubric-2026.11.4 — the same rubric and the same method as every other entry in this index.
  • Measured by Canine Code Analyzer (cca.canine.dev) using cca measure-2026.11.73+preprod-52de90b74216.