You NEED to try these 12 open-source AI projects RIGHT NOW
Open-weights vision language model from Baidu for fast OCR and document analysis, shown highlighting a PDF. The model is about 6.5 GB and was released a couple of days earlier.
Baidu's open-weights vision-language OCR model for parsing very long, multi-page documents.
Updated EU status checked Last shown
Unlimited OCR at a glance
- What is Unlimited OCR?
- Unlimited OCR is a model from Baidu. Baidu's open-weights vision-language OCR model for parsing very long, multi-page documents.
- Who makes Unlimited OCR?
- Unlimited OCR is made by Baidu.
- Does Unlimited OCR work in the EU?
- Available in the EU. Open weights are downloadable from Hugging Face and ModelScope and can be self-hosted anywhere, and no EU restriction appears in the provided material. The license text was not seen, so this is not confirmed. Last verified 14 Nov 2026.
- How much does Unlimited OCR cost?
- Open source. Weights downloadable for free. Also offered through Baidu Cloud OCR; pricing there was not provided.
- Where was Unlimited OCR shown?
- In 1 creator video by Matthew Berman. Most recently by Matthew Berman on 24 Jun 2026, at 10:11 in “You NEED to try these 12 open-source AI projects RIGHT NOW”.
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
About
Unlimited OCR is a document-parsing vision-language model from Baidu, positioned as a follow-up to DeepSeek-OCR. The first version was released in June 2026. Unlimited OCR 2, released on 2026-10-27, adds Sparse Reference Sliding Window Attention (SR-SWA). SR-SWA compresses text-visual attention to a fixed-size set of working references (e.g. 512 tokens), so per-step decoding latency and memory stay nearly constant as documents grow.
The model supports single-page parsing with dynamic tiling and multi-page parsing, where all pages are passed in one request so the model can use cross-page context. Baidu reports state-of-the-art results on MPDocBench and on its own MarathonDocBench, a benchmark for hundred-page documents. These are vendor claims.
It runs with Hugging Face Transformers for quick tests, or with a customized SGLang build recommended for serving. vLLM support was added for version 1, and ms-swift supports training. The README tests on an NVIDIA A800-80GB. It is also listed on ModelScope and Baidu Cloud, with a Hugging Face Spaces demo. A YouTube reviewer described the weights as about 6.5 GB. The license was not stated in the provided material.
Key features
- Parses multi-page documents in a single request with cross-page context
- SR-SWA keeps decoding latency and memory near constant for long documents
- Single-page mode with dynamic tiling (gundam mode)
- Inference via Transformers, SGLang and vLLM (v1); training via ms-swift
- Page-window scheduling for very long documents in SGLang
- Companion MarathonDocBench benchmark for ultra-long document parsing
1 moment
As seen on YouTube
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Keep exploring
Similar AI tools
Closest neighbours by meaning, not just by category.
UO
Unlimited OCR 2
Model · Baidu
Baidu's open-source OCR model that parses very long multi-page documents in one pass with near-constant decoding cost.
Available in EUPlatforms & Infra
27 Oct 2026Open source
E5
ERNIE 5.5 Flash
Model · Baidu
Baidu's open-weights 280B-parameter sparse MoE base model (14B active) with 1M-token context, released pre-post-training.
Available in EULanguage Models
5 Nov 2026Open source
DN
Dots3 Note
Model · Rednote
Open MoE language model from Rednote (Xiaohongshu) with 208B total and 16B active parameters.
UnverifiedLanguage Models
Release date n/a
NN
NVIDIA Nemotron 3.5 Super VL
Model · NVIDIA
NVIDIA's open multimodal model (120B total, 12B active parameters) for image, video and document understanding.
UnverifiedLanguage Models
Release date n/aOpen source
DV
DeepSeek V4.1 Flash
Model · DeepSeek
Open-weights multimodal DeepSeek model in the V4 family, offered via API, chat and downloadable weights.
Available in EULanguage Models
Release date n/aOpen source
M4
MiniCPM-V 4.6
Model · OpenBMB
Small open-weights multimodal model from OpenBMB for image and video understanding on phones and edge devices.
Available in EULanguage Models
13 Apr 2026Open source