Releases · PaddlePaddle/PaddleOCR (original) (raw)

v3.7.0

2026.6.11 v3.7.0 released

• Release PP-OCRv6
  • Accuracy boost: Medium tier achieves +4.6% detection and +5.1% recognition over PP-OCRv5_server, surpassing mainstream VLMs (Qwen3-VL-235B, GPT-5.5) with only 34.5M parameters.
  • 50 languages unified: Single model covers Chinese, English, Japanese, and 46 Latin-script languages — no model switching needed.
  • Specialized scenarios: Major improvements in digital displays, dot-matrix characters, tire prints, and industrial text recognition.
  • Faster inference: 5.2× CPU speedup (OpenVINO), 6.1× on Apple M4 (tiny), 0.13s on A100 GPU.
  • Three tiers for all scenarios: tiny (1.5M) / small (7.7M) / medium (34.5M) for edge, mobile, and server deployment.

2026.6.11 v3.7.0 发布

• 发布 PP-OCRv6
  • 精度全面提升：medium 档相比 PP-OCRv5_server 检测精度提升 4.6%、识别精度提升 5.1%，以仅 34.5M 参数超越 Qwen3-VL-235B、GPT-5.5 等主流视觉语言大模型。
  • 50 种语言统一支持：单一模型覆盖中文、英文、日文及 46 种拉丁语系语言，无需为不同语种切换模型。
  • 专业场景增强：数码显示屏、点阵字符、轮胎印字、工业字符等传统 VLM 难以覆盖的场景识别能力大幅提升。
  • 推理速度更快：medium 档 CPU OpenVINO 推理加速 5.2×，tiny 档 Apple M4 加速 6.1×，A100 上仅需 0.13s。
  • 三档模型覆盖全场景：tiny（1.5M）/ small（7.7M）/ medium（34.5M）分别面向端侧/移动端/服务端部署。

Full Changelog: v3.6.0...v3.7.0

v3.6.0

2026.5.28 v3.6.0 released

• Release the PaddleOCR-VL-1.6 document parsing solution.
  • PaddleOCR-VL-1.6 is an upgraded compact document parsing model built upon PaddleOCR-VL-1.5. Through a newly introduced region-aware data optimization framework and a progressive post-training recipe, the model further enhances parsing robustness and overall recognition accuracy.
  • PaddleOCR-VL-1.6 achieves a new state-of-the-art score of 96.33% on the authoritative benchmark OmniDocBench v1.6, while also setting new records on OmniDocBench v1.5 and Real5-OmniDocBench. The model delivers leading performance across text, formula, and table recognition tasks, surpassing both open-source and closed-source solutions.
  • PaddleOCR-VL-1.6 significantly improves capabilities in table recognition, Chinese ancient document parsing, and Chinese rare character recognition, while further enhancing performance in seal recognition, text spotting, chart recognition, and other diverse document scenarios.
  • The model architecture remains fully compatible with PaddleOCR-VL-1.5, enabling zero-cost plug-and-play migration for existing users.
• Release official PaddleOCR API SDKs for Python, Go, and TypeScript, enabling more convenient integration with PaddleOCR official asynchronous APIs.
• Add support for parsing multi-page TIFF files.

2026.5.28 v3.6.0 发布

• 发布 PaddleOCR-VL-1.6 文档解析方案。
  • PaddleOCR-VL-1.6 是基于 PaddleOCR-VL-1.5 全新升级的轻量级文档解析模型。通过引入全新的区域感知数据优化框架以及渐进式后训练优化方案，模型进一步提升了复杂文档场景下的解析鲁棒性与整体识别精度。
  • PaddleOCR-VL-1.6 在权威文档解析评测集 OmniDocBench v1.6 上取得了 96.33% 的全新 SOTA 成绩，并同时刷新了 OmniDocBench v1.5 与 Real5-OmniDocBench 的最佳纪录。在文本、公式、表格等核心任务上，模型均展现出全面领先的性能，超越当前主流开源及闭源方案。
  • PaddleOCR-VL-1.6 在表格识别、中文古籍文档解析、中文生僻字识别等能力上实现显著增强，同时进一步提升了印章识别、文本检测识别（Spotting）、图表解析等多种复杂场景下的表现。
  • PaddleOCR-VL-1.6 与 PaddleOCR-VL-1.5 保持完全兼容，无需额外适配即可实现零成本即插即用升级。
• 发布 PaddleOCR 官方 API Python、Go、TypeScript SDK，支持更加便捷地调用 PaddleOCR 官方异步 API。
• 新增对多页 TIFF 文件解析的支持。

Full Changelog: v3.5.0...v3.6.0

v3.5.0

2026.4.21 v3.5.0 released

• Deeply integrated with the Hugging Face ecosystem, with 20 major models supporting Transformers as the inference backend. Supports flexible switching of inference engines, including PaddlePaddle static graph, PaddlePaddle dynamic graph, or Transformers.
• Supports conversion of common document formats (Word, Excel, Powerpoint) to Markdown.
• The PaddleOCR-VL series, PP-StructureV3, and PP-DocTranslation support exporting parsing results to DOCX format, making it convenient to view and edit in Word.
• Official browser inference SDK PaddleOCR.js is released, supporting running PP-OCRv5 in the browser.

2026.4.21 v3.5.0 发布

• 深度适配 Hugging Face 生态，20 个主要模型支持以 Transformers 作为推理后端。支持灵活切换推理引擎，可选飞桨静态图、飞桨动态图或 Transformers。
• 支持常见文档格式（Word、Excel、Powerpoint）转 Markdown。
• PaddleOCR-VL 系列、PP-StructureV3、PP-DocTranslation 支持将解析结果导出为 DOCX 格式，便于在 Word 中查看和编辑。
• 发布官方浏览器推理 SDK PaddleOCR.js，支持在浏览器中运行 PP-OCRv5。

Full Changelog: v3.4.1...v3.5.0

v3.4.1

2026.4.14 v3.4.1 released

• PaddleOCR-VL adds llama-cpp-server backend support.
• PaddleOCR-VL adds AMD GPU and Intel Arc GPU hardware support.
• Fixed dependency issues in the PaddleOCR-VL images for Huawei NPU and KunlunXin XPU*
• For the PaddleOCR-VL Docker Compose service, the default configuration no longer limits the maximum number of pages per request.

2026.4.14 v3.4.1 发布

• PaddleOCR-VL 新增 llama-cpp-server 后端支持。
• PaddleOCR-VL 新增 AMD GPU、Intel Arc GPU 硬件支持。
• 修复 PaddleOCR-VL 华为 NPU、昆仑芯 XPU 镜像中的依赖问题。
• 对于 PaddleOCR-VL Docker Compose 服务，默认不限制单请求的页数上限。

Full Changelog: v3.4.0...v3.4.1

v3.4.0

2026.1.29 v3.4.0 released

• Release the PaddleOCR-VL-1.5 complex document parsing solution.
  PaddleOCR-VL-1.5 is a new iterative version of the PaddleOCR-VL series. Based on comprehensive optimization of the core capabilities of version 1.0, the model achieves 94.5% accuracy on the authoritative document parsing benchmark OmniDocBench v1.5, surpassing top global general-purpose large models and document parsing–specific models.
  PaddleOCR-VL-1.5 innovatively supports irregular-shaped bounding box localization of document elements, enabling excellent performance in real-world application scenarios such as scanning, skew, warping, screen-photography, and complex illumination, achieving comprehensive SOTA performance. In addition, the model further integrates seal recognition and spotting tasks, with key metrics continuing to lead mainstream models.
  You can use it online on the PaddleOCR official website or call the model API.
• Add support for calling MLX-VLM inference services.
• PaddleOCR-VL now supports cross-page table merging and multi-level heading reconstruction.
• PP-StructureV3 adds support for the format_block_content and markdown_ignore_labels parameters.
• Fixed an issue where accessing the /docs endpoint in the official PaddleOCR-VL image would result in an error.

2026.1.29 v3.4.0 发布

• 发布 PaddleOCR-VL-1.5 复杂文档解析方案。
  PaddleOCR-VL-1.5 是 PaddleOCR-VL 系列的全新迭代版本。在全面优化 1.0 版本核心能力的基础上，该模型在文档解析权威评测集 OmniDocBench v1.5 上斩获了 94.5% 的高精度，超越了全球的顶尖通用大模型及文档解析专用模型。
  PaddleOCR-VL-1.5 创新性地支持了文档元素的异形框定位，使得 PaddleOCR-VL-1.5 在扫描、倾斜、弯折、屏幕拍摄及复杂光照等真实落地场景中均表现卓越，实现了全面的 SOTA。此外，模型进一步集成了印章识别与文本检测识别任务，关键指标持续领跑主流模型。
  您可以在 PaddleOCR官网 在线使用或者调用该模型的API。
• 新增对 MLX-VLM 推理服务的调用支持。
• PaddleOCR-VL 支持合并跨页表格、多级标题重建功能。
• PP-StructureV3 支持 format_block_content、markdown_ignore_labels 参数。
• 修复 PaddleOCR-VL 官方镜像访问 /docs 接口报错的问题。

New Contributors

• @AmirHosseinOmidi0 made their first contribution in #16659
• @ZhangX-21 made their first contribution in #16745
• @AdlerFleurant made their first contribution in #16756
• @tianyuzhou668 made their first contribution in #16518
• @shiyuasuka made their first contribution in #17041
• @1250890838 made their first contribution in #16996
• @Ihebdhouibi made their first contribution in #16994
• @Ghazi-raad made their first contribution in #17201
• @orbisai0security made their first contribution in #17289
• @danghoangnhan made their first contribution in #17019
• @Luxorion-12 made their first contribution in #17158

Full Changelog: v3.3.3...v3.4.0

v3.3.3

2026.1.20 v3.3.3 released

• PaddleOCR-VL now supports specifying custom model names and API keys, and can seamlessly integrate with inference services from third-party platforms such as SiliconFlow and Novita AI.
• The PP-StructureV3 MCP Server supports using hosted services on the Qianfan platform as the underlying inference engine.
• The documentation for PP-OCRv5 and PaddleOCR-VL has been comprehensively improved, with known errors fixed to enhance readability and accuracy.
• Added support for inference on Muxi GPUs, further expanding hardware compatibility and deployment flexibility.

2026.1.20 v3.3.3 发布

• PaddleOCR-VL 现已支持指定自定义模型名称与 API Key，并可无缝对接硅基流动、Novita AI 等第三方平台的推理服务。
• PP-StructureV3 MCP Server 支持基于千帆平台的托管服务作为底层推理引擎。
• 全面优化 PP-OCRv5 与 PaddleOCR-VL 相关文档，修复已知错漏，提升可读性与准确性。
• 新增对沐曦 GPU 的推理支持，进一步扩展硬件兼容性与部署灵活性。

New Contributors

• @metax666 made their first contribution in #17269

Full Changelog: v3.3.2...v3.3.3

v3.3.2

v3.3.1

2025.10.29 v3.3.1 released

• Fixed the issue where the document image preprocessing switch did not take effect in PP-StructureV3 and PaddleOCR-VL.
• Improved the user documentation by adding instructions for using PaddleOCR-VL in a fully offline environment. Details.

2025.10.29 v3.3.1 发布

• 修复了 PP-StructureV3 和 PaddleOCR-VL 的文档图像预处理开关不生效的问题。
• 优化了使用文档，提供了 PaddleOCR-VL 纯离线环境使用的方式。详情。

New Contributors

• @AmirHosseinOmidi0 made their first contribution in #16659
• @ZhangX-21 made their first contribution in #16745
• @AdlerFleurant made their first contribution in #16756
• @tianyuzhou668 made their first contribution in #16518

Full Changelog: v3.3.0...v3.3.1

v3.3.0

2025.10.16 v3.3.0 released

• Released PaddleOCR-VL:
  • Model Introduction:
    * PaddleOCR-VL is a SOTA and resource-efficient model tailored for document parsing. Its core component is PaddleOCR-VL-0.9B, a compact yet powerful vision-language model (VLM) that integrates a NaViT-style dynamic resolution visual encoder with the ERNIE-4.5-0.3B language model to enable accurate element recognition. This innovative model efficiently supports 109 languages and excels in recognizing complex elements (e.g., text, tables, formulas, and charts), while maintaining minimal resource consumption. Through comprehensive evaluations on widely used public benchmarks and in-house benchmarks, PaddleOCR-VL achieves SOTA performance in both page-level document parsing and element-level recognition. It significantly outperforms existing solutions, exhibits strong competitiveness against top-tier VLMs, and delivers fast inference speeds. These strengths make it highly suitable for practical deployment in real-world scenarios. The model has been released on HuggingFace. Everyone is welcome to download and use it!
  • Core Features:
    * Compact yet Powerful VLM Architecture: We present a novel vision-language model that is specifically designed for resource-efficient inference, achieving outstanding performance in element recognition. By integrating a NaViT-style dynamic high-resolution visual encoder with the lightweight ERNIE-4.5-0.3B language model, we significantly enhance the model’s recognition capabilities and decoding efficiency. This integration maintains high accuracy while reducing computational demands, making it well-suited for efficient and practical document processing applications.
    * SOTA Performance on Document Parsing: PaddleOCR-VL achieves state-of-the-art performance in both page-level document parsing and element-level recognition. It significantly outperforms existing pipeline-based solutions and exhibiting strong competitiveness against leading vision-language models (VLMs) in document parsing. Moreover, it excels in recognizing complex document elements, such as text, tables, formulas, and charts, making it suitable for a wide range of challenging content types, including handwritten text and historical documents. This makes it highly versatile and suitable for a wide range of document types and scenarios.
    * Multilingual Support: PaddleOCR-VL Supports 109 languages, covering major global languages, including but not limited to Chinese, English, Japanese, Latin, and Korean, as well as languages with different scripts and structures, such as Russian (Cyrillic script), Arabic, Hindi (Devanagari script), and Thai. This broad language coverage substantially enhances the applicability of our system to multilingual and globalized document processing scenarios.
• Released PP-OCRv5 Multilingual Recognition Model:
  • Improved the accuracy and coverage of Latin script recognition; added support for Cyrillic, Arabic, Devanagari, Telugu, Tamil, and other language systems, covering recognition of 109 languages. The model has only 2M parameters, and the accuracy of some models has increased by over 40% compared to the previous generation.

2025.10.16 v3.3.0 发布

• 发布PaddleOCR-VL：
  • 模型介绍：
    * PaddleOCR-VL 是一款先进、高效的文档解析模型，专为文档中的元素识别设计。其核心组件为 PaddleOCR-VL-0.9B，这是一种紧凑而强大的视觉语言模型（VLM），它由 NaViT 风格的动态分辨率视觉编码器与 ERNIE-4.5-0.3B 语言模型组成，能够实现精准的元素识别。该模型支持 109 种语言，并在识别复杂元素（如文本、表格、公式和图表）方面表现出色，同时保持极低的资源消耗。通过在广泛使用的公开基准与内部基准上的全面评测，PaddleOCR-VL 在页级级文档解析与元素级识别均达到 SOTA 表现。它显著优于现有的基于Pipeline方案和文档解析多模态方案以及先进的通用多模态大模型，并具备更快的推理速度。这些优势使其非常适合在真实场景中落地部署。模型已发布至HuggingFace，欢迎大家下载使用！
  • 特性：
    * 紧凑而强大的视觉语言模型架构：我们提出了一种新的视觉语言模型，专为资源高效的推理而设计，在元素识别方面表现出色。通过将NaViT风格的动态高分辨率视觉编码器与轻量级的ERNIE-4.5-0.3B语言模型结合，我们显著增强了模型的识别能力和解码效率。这种集成在保持高准确率的同时降低了计算需求，使其非常适合高效且实用的文档处理应用。
    * 文档解析的SOTA性能：PaddleOCR-VL在页面级文档解析和元素级识别中达到了最先进的性能。它显著优于现有的基于流水线的解决方案，并在文档解析中展现出与领先的视觉语言模型（VLMs）竞争的强劲实力。此外，它在识别复杂的文档元素（如文本、表格、公式和图表）方面表现出色，使其适用于包括手写文本和历史文献在内的各种具有挑战性的内容类型。这使得它具有高度的多功能性，适用于广泛的文档类型和场景。
    * 多语言支持：PaddleOCR-VL支持109种语言，覆盖了主要的全球语言，包括但不限于中文、英文、日文、拉丁文和韩文，以及使用不同文字和结构的语言，如俄语（西里尔字母）、阿拉伯语、印地语（天城文）和泰语。这种广泛的语言覆盖大大增强了我们系统在多语言和全球化文档处理场景中的适用性。
• 发布PP-OCRv5小语种识别模型：
  • 优化拉丁文识别的准度和广度，新增西里尔文、阿拉伯文、天城文、泰卢固语、泰米尔语等语系，覆盖109种语言文字的识别。模型参数量仅为2M，部分模型精度较上一代提升40%以上。

v3.2.0

2025.8.21 v3.2.0 released

• Significant Model Additions:
  • Introduced training, inference, and deployment for PP-OCRv5 recognition models in English, Thai, and Greek. The PP-OCRv5 English model delivers an 11% improvement in English scenarios compared to the main PP-OCRv5 model, with the Thai and Greek recognition models achieving accuracies of 82.68% and 89.28%, respectively.
• Deployment Capability Upgrades:
  • Full support for PaddlePaddle framework versions 3.1.0 and 3.1.1.
  • Comprehensive upgrade of the PP-OCRv5 C++ local deployment solution, now supporting both Linux and Windows, with feature parity and identical accuracy to the Python implementation.
  • High-performance inference now supports CUDA 12, and inference can be performed using either the Paddle Inference or ONNX Runtime backends.
  • The high-stability service-oriented deployment solution is now fully open-sourced, allowing users to customize Docker images and SDKs as required.
  • The high-stability service-oriented deployment solution also supports invocation via manually constructed HTTP requests, enabling client-side code development in any programming language.
• Benchmark Support:
  • All production lines now support fine-grained benchmarking, enabling measurement of end-to-end inference time as well as per-layer and per-module latency data to assist with performance analysis.
  • Documentation has been updated to include key metrics for commonly used configurations on mainstream hardware, such as inference latency and memory usage, providing deployment references for users.
• Bug Fixes:
  • Resolved the issue of failed log saving during model training.
  • Upgraded the data augmentation component for formula models for compatibility with newer versions of the albumentations dependency, and fixed deadlock warnings when using the tokenizers package in multi-process scenarios.
  • Fixed inconsistencies in switch behaviors (e.g., use_chart_parsing) in the PP-StructureV3 configuration files compared to other pipelines.
• Other Enhancements:
  • Separated core and optional dependencies. Only minimal core dependencies are required for basic text recognition; additional dependencies for document parsing and information extraction can be installed as needed.
  • Enabled support for NVIDIA RTX 50 series graphics cards on Windows; users can refer to the installation guide for the corresponding PaddlePaddle framework versions.
  • PP-OCR series models now support returning single-character coordinates.
  • Added AIStudio, ModelScope, and other model download sources, allowing users to specify the source for model downloads.
  • Added support for chart-to-table conversion via the PP-Chart2Table module.
  • Optimized documentation descriptions to improve usability.

2025.8.21 v3.2.0 发布

• 重要模型新增：
  • 新增 PP-OCRv5 英文、泰文、希腊文识别模型的训练、推理、部署。其中 PP-OCRv5 英文模型较 PP-OCRv5 主模型在英文场景提升 11%，泰文识别模型精度 82.68%，希腊文识别模型精度 89.28%。
• 部署能力升级：
  • 全面支持飞桨框架 3.1.0 和 3.1.1 版本。
  • 全面升级 PP-OCRv5 C++ 本地部署方案，支持 Linux、Windows，功能及精度效果与 Python 方案保持一致。
  • 高性能推理支持 CUDA 12，可使用 Paddle Inference、ONNX Runtime 后端推理。
  • 高稳定性服务化部署方案全面开源，支持用户根据需求对 Docker 镜像和 SDK 进行定制化修改。
  • 高稳定性服务化部署方案支持通过手动构造HTTP请求的方式调用，该方式允许客户端代码使用任意编程语言编写。
• Benchmark支持：
  • 全部产线支持产线细粒度 benchmark，能够测量产线端到端推理时间以及逐层、逐模块的耗时数据，可用于辅助产线性能分析。
  • 文档中补充各产线常用配置在主流硬件上的关键指标，包括推理耗时和内存占用等，为用户部署提供参考。
• Bug修复：
  • 修复模型训练时训练日志保存失败的问题。
  • 对公式模型的数据增强部分进行了版本兼容性升级，以适应新版本的 albumentations 依赖，并修复了在多进程使用 tokenizers 依赖包时出现的死锁警告。
  • 修复 PP-StructureV3 配置文件中的 use_chart_parsing 等开关行为与其他产线不统一的问题。
• 其他升级：
  • 分离必要依赖与可选依赖。使用基础文字识别功能时，仅需安装少量核心依赖；若需文档解析、信息抽取等功能，用户可按需选择安装额外依赖。
  • 支持 Windows 用户使用英伟达 50 系显卡，可根据安装文档安装对应版本的 paddle 框架。
  • PP-OCR 系列模型支持返回单文字坐标。
  • 模型新增 AIStudio、ModelScope 等下载源。可指定相关下载源下载对应的模型。
  • 支持图表转表 PP-Chart2Table 单功能模块推理能力。
  • 优化部分使用文档中的描述，提升易用性。

New Contributors

• @awkrail made their first contribution in #16262
• @liuwake made their first contribution in #16281

Full Changelog: v3.1.1...v3.2.0