Understand this story
On October 27, 2026, Baidu released and open-sourced the document parsing model Unlimited OCR 2. According to Baidu's official introduction, the model uses a new SR-SWA mechanism to compress the attention interaction between text and vision into a fixed-size working reference, thereby parsing documents of almost arbitrary length within a limited context window; Baidu says that as documents grow longer, per-step decoding latency and memory usage remain nearly constant, and that it achieves SOTA performance on MPDocBench and MarathonDocBench, which cover documents ranging from dozens to hundreds of pages. On the same day, QbitAI reported that in a single-sequence test at the hundred-page level, the model reduced peak KV cache from 47GB under standard decoding to 1.77GB, a reduction of about 96%; in a 16-way concurrent test, end-to-end throughput rose from 360 to 1315 tokens/second, about 3.65 times that of standard decoding. The model has been open-sourced on Hugging Face and GitHub.
AI-generated from public reports
Latest update · Discussion
Shao Meng joined this discussionDevelopments
1 confirmed developments
- Source:English text is being prepared.
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Oct 27
Shao MengRelated discussion
Shao Meng joined this discussion
This post contributes to heat; it is not a formal report or confirmed fact.
Shao Meng
Baidu Open-Sources Unlimited OCR 2: 3B/A0.5B MoE Architecture Supports Ultra-Long Document Parsing
Baidu has open-sourced Unlimited OCR 2, an OCR and layout parsing model designed for documents ranging from dozens to hundreds of pages, adopting a 3B/A0.5B MoE architecture. Compared with v1's R-SWA, v2 upgrades the decoder attention to SR-SWA and adds Unlimited-Horizon Decoding, which compresses the attention interaction between text and vision into a fixed-capacity working reference set (the README gives 512 tokens as an example), and then combines this with page-by-page chunked decoding, enabling ultra-long documents to be parsed within a limited context. The authors claim that at the hundred-page scale, step-by-step decoding speed is approximately 7.0× that of Full Attention and approximately 5.6× that of R-SWA, with latency and GPU memory barely increasing as documents grow longer.
Adina YakupRelated discussion
Adina Yakup joined this discussion
This post contributes to heat; it is not a formal report or confirmed fact.
Berry XiaRelated discussion
Berry Xia joined this discussion
This post contributes to heat; it is not a formal report or confirmed fact.
QbitAI
Baidu Open-Sources ERNIE Unlimited OCR 2, Peak KV Cache Reduced from 47GB to 1.77GB
Baidu has released and open-sourced ERNIE Unlimited OCR 2. Through the sparse reference sliding window attention mechanism SR-SWA and an infinite decoding mechanism, it reduced the peak KV cache from 47GB under standard decoding to 1.77GB in a single-sequence test at the hundred-page level, a reduction of about 96%. In a 16-way concurrency test, end-to-end throughput increased from 360 to 1315 tokens/second, about 3.65 times that of standard decoding. The model has been open-sourced on Hugging Face and GitHub; the previous generation, Unlimited OCR, has accumulated more than 10 million downloads and about 28,600 GitHub stars.
Leiphone
Baidu Open-Sources ERNIE Unlimited OCR 2, Supporting Document Parsing of Any Page Count
Baidu has released and open-sourced ERNIE Unlimited OCR 2, claiming it is the first to achieve document parsing of any page count; in testing, it parsed all 7 books of the Harry Potter series in a single pass, totaling more than 3,600 pages. The model achieved overall scores of 84.6 and 77.1 on MPDocBench and MarathonDocBench, respectively. It adopts the Sparse Reference Sliding Window Attention (SR-SWA) mechanism, and in hundred-page document scenarios, its single-step decoding speed is 7 times and 5.6 times faster than Full Attention and the previous-generation R-SWA, respectively. The code and model weights have been open-sourced on GitHub and Hugging Face, and the model is also available on the Baidu AI Cloud platform and the Baidu AI Open Platform.
MarkTechPost
Baidu Open-Sources Unlimited OCR 2, a 3.34B Model Supporting Parsing of Hundreds of Pages of Documents
Baidu has open-sourced Unlimited OCR 2, a 3.34B-parameter vision-language model that can parse hundreds of pages of books, financial reports, and reports in a single request. The weights are released under the MIT license, and the code is on GitHub. Its core change is SR-SWA sparse attention, which compresses text and visual attention into a fixed 512-token reference set, keeping the latency and memory of each decoding step essentially flat as documents grow longer. Baidu self-reports that the model scores 84.6 on MPDocBench and 77.1 on MarathonDocBench, where the next-best model Kimi-K3 scores 33.7, though its heading-hierarchy sub-item is only 48.2.
BaiduFirst-partySelected
Baidu Open-Sources Unlimited OCR 2 Document Parsing Model
Baidu has open-sourced Unlimited OCR 2, which uses a new SR-SWA mechanism to compress the attention interaction between text and vision into a fixed-size working reference, thereby parsing documents of almost any length within a limited context window. Baidu says that as documents grow longer, per-step decoding latency and memory usage remain nearly constant, and that the model achieves SOTA performance on MPDocBench and MarathonDocBench, which cover documents ranging from dozens to hundreds of pages.
Heat trend for this story
- Current
- 36
- Period peak
- 53 Oct 27 20:00
- 24-hour change
- +96%
Move the pointer or select the chart to inspect hourly heat. Use the left and right arrow keys with a keyboard.