# Tencent-Hunyuan/HunyuanOCR

HunyuanOCR-1.5: Making Lightweight OCR VLMs Faster and Better

Repository: https://github.com/Tencent-Hunyuan/HunyuanOCR
Canonical: https://ross.abutalabs.com/products/hunyuanocr
Language: Python
License: NOASSERTION
License Family: other
Last push: 2026-07-29T06:29:52+00:00

## Health v2 (maintenance only)
Score: 59/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 95, release rhythm 35, longevity 20
- inputs: {"age_days": 288, "days_push": 35, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1930, forks 150 (observed 2026-08-28T04:05:56.155002+00:00)

## What it is
HunyuanOCR-1.5 is a lightweight end-to-end OCR vision-language model from Tencent, with a unified inference environment, llama.cpp PC-side deployment, and a verl-based reinforcement learning training stack. It also ships related benchmarks such as CHAOS-Bench for character-level hallucination probing.

## Use cases
- extract text from images and scanned documents
- parse document layouts including tables
- run a fast lightweight OCR model on a PC
- translate text inside images
- benchmark OCR models for character-level hallucination
- fine-tune an OCR vision-language model with reinforcement learning

## When to choose
- you need an open-source OCR VLM with document, table, and chart parsing
- you want lightweight on-device or PC-side OCR inference via llama.cpp
- you need a trainable OCR model with an RL training stack

## When to avoid
- you only need simple Tesseract-style text extraction without a GPU or VLM
- you need a fully permissive license (license is custom/NOASSERTION)
- you cannot run GPU inference or large model weights

## Facets
- artifact type: library
- maturity: active
- function: ocr, machine-learning, llm-inference, deep-learning
- domain: computer-vision, artificial-intelligence, pdf
- platform: python, cross-platform
- tags: vision-language-model, document-parsing, speculative-decoding, llama.cpp, reinforcement-learning, tencent-hunyuan, table-parsing, benchmarks, natural-language-processing, gpu, linux

## Member repositories
- Tencent-Hunyuan/HunyuanOCR (main) score 59

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:56.155002+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:08:46.886792+00:00, confidence not recorded.
  - readme: https://github.com/Tencent-Hunyuan/HunyuanOCR (fetched 2026-08-28T04:05:56.155002+00:00, sha cceb7af4cfff)
- Data as of 2026-08-30T08:39:29.467469+00:00.