Skip to content
Computer vision

MiniCPM-OCR 1.0

компактная OCR-модель для документов

1 min read

MiniCPM-OCR 1.0 (1.3B) — первая OCR-специализированная модель в семействе MiniCPM-V/o от OpenBMB. Обучена на базе MiniCPM-V 4.6 для разбора текста, таблиц, формул, графиков, печатей и извлечения структурированной информации; основные языки — китайский и английский.

Ключевые особенности

  • SOTA-разбор документов: 93.68 на Real5-OmniDocBench и лидерство на OmniDocBench v1.6, Wild-OmniDocBench и olmOCR-Bench.
  • Ускорение DSpark: до +55% к end-to-end эффективности; пропускная способность ~1.49× выше PaddleOCR-VL-1.6 и ~3.62× выше MinerU2.5 Pro.
  • Понимает текст, таблицы, формулы, графики и печати; умеет KIE с выдачей JSON.
  • Готовность к деплою: vLLM, SGLang, llama.cpp и Ollama с первого дня, квантованные варианты и мультичиповые сборки.
  • Apache 2.0, pip-SDK, API, агентные workflow и возможности дообучения.

Пример

После запуска vLLM-сервиса с моделью можно разобрать изображение через CLI:

pip install "minicpm-ocr[all]"
minicpmvocr parse demo.png --backend vllm --server-url http://127.0.0.1:8080 --model minicpm-ocr --device cuda:0 --output output/demo.md
Подробнее Бенчмарки: 93.68 на Real5-OmniDocBench, 54.7 на OCRBench v2; ~1.2 страницы/с на RTX 5090 при concurrency 1.
Пайплайн: MiniCPM-OCR 1.0 + PP-DocLayoutV3 + SDK minicpmvocr; визуальные токены с 4× сжатием, scale_resolution 448, до 9 слайсов.
Ограничения: основные языки — китайский и английский; для SDK нужен отдельно запущенный сервис модели.
Лицензия Apache 2.0; есть BF16, GGUF, AWQ W4A16 и GPTQ W4A16 варианты.
Примеры и видео
Discuss Telegram 416 10.4K

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 2.4 GB · ready files3.2 GB
  • FP8 / INT8recommendedweights 1.2 GB1.8 GB
  • 4-bitweights 0.7 GB1.2 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # openbmb/MiniCPM-OCR-1.0
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download openbmb/MiniCPM-OCR-1.0 --local-dir ~/models/MiniCPM-OCR-1.0
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to MiniCPM-OCR-1.0
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of MiniCPM-V-4.6 (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 8All on Hugging Face

About the model

openbmb/MiniCPM-OCR-1.0
Author
openbmb
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

10.4K416

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments

Language models

PRISM-1B

PRISM-1B