HunyuanOCR

Commercial-grade, lightweight (1B) end-to-end OCR expert VLM. Connects a 0.4B native-resolution ViT to a 0.5B Hunyuan LLM. Won first place in ICDAR 2025 DIMT Challenge (Small Model Track).

HunyuanOCR-1.5 (paper July 6, 2026, arXiv 2607.04884) succeeds this model with a reported ~6.4x decoding speedup, ancient-script and chart-parsing support, and broader multilingual coverage; weights not yet released at paper time.