Open Source Project

Ant Group has launched a new-generation spatial perception model for robots that enhances 3D depth perception capabilities and improves performance in grasping, navigation, and environmental understanding.

An open-source, pure front-end GUI agent framework from Alibaba that requires no screenshots or backend; with just one line of code, it enables any web page to support natural language interaction and automated operations.

Alibaba's open-source large-scale speech model supports zero-shot cloning in 3 seconds, multilingual capabilities, and command-based emotional control, enabling ultra-low-latency streaming synthesis at 150 ms.

A Microsoft open-source framework that automatically optimizes agent skills—just like training a neural network—without fine-tuning the model, enabling consistent improvement in capabilities and cross-environment transfer.

AliQianwen, the first native language world model released on June 24, 2026, uses plain text to uniformly simulate seven major digital environments, allowing agents to practice in a "virtual world."

Alibaba's open-source AI code review CLI tool employs a hybrid architecture combining "deterministic engineering" and an "LLM agent." It performs line-level code defect detection, supports multiple models, and can be integrated into CI/CD pipelines. It is fully open-sourced under the Apache 2.0 license.

An open-source project based on Claude Code that uses 48 layered AI agents to simulate an entire game development team, enabling a single person to manage the entire process from design to launch.

An open-source AI short-form drama production platform—a fully automated pipeline that takes a novel and turns it into a finished video. It features skeleton binding to prevent face distortion, produces an episode in 8 minutes, and keeps costs down to just over ten yuan.

An extreme Token-saving, open-source, general-purpose AI Agent with Skill skill ecosystem support that automates programming, office and all kinds of complex tasks for you locally at a very low cost.

A 4.2 billion-parameter open source body intelligence model developed by Variable Robotics realizes “out-of-the-box” zero-sample deployment capability by virtue of its innovative end-to-end architecture, allowing developers to empower robots with powerful cognitive, reasoning, and fine manipulation capabilities with only a consumer-grade graphics card.

The V4 model series, developed through in-depth research, includes V4.1-Flash—which supports text-and-image understanding, long context, and tool invocation—as well as V4 Pro, which continues to provide services.

The world's first open-source quantum AI model series, through AI-driven quantum chip calibration and error correction, provides a high-performance tool chain for practical quantum computing and reshapes the quantum industry ecosystem.

The 2026 open source AI video generation benchmark, with a single-stream Transformer architecture to achieve text/image to 1080p HD video generation at breakneck speeds, and native support for multi-language lip-synchronization and sound generation, topped the global performance list.

Open source and free AI video translation and dubbing tool, supporting multi-language speech recognition, subtitle translation and natural dubbing, helping content creators and enterprises to easily realize the globalization of video distribution.

Open source AI public opinion tool, multi-agent collaboration to analyze the whole network data, can accurately insight into the trend, predict the direction, applicable to brand public relations, market research and other scenarios.

Mistral AI introduces an open source, low-latency text-to-speech model that supports cross-language timbre cloning with latency as low as 70ms and can be deployed at the edge.

An open source AI intelligence framework for local file management, cross-tool automation, and lightweight development assistance via natural language commands, balancing privacy protection with low-code ease of use.

Google's open source lightweight multimodal translation model supports 55 languages and image translations, with performance that exceeds larger models, taking into account both mobile and cloud deployments, and facilitating efficient globalized communication.

Meta introduces the world's first unified multimodal audio separation model that supports text, visual, and time cues to accurately separate target sounds from complex audio and video.

An AI tool developed by Peking University can automatically convert papers and text into editable PowerPoint presentations and structural diagrams. Supporting multimodal input, it efficiently addresses the challenges of scientific diagramming and converting lengthy documents into reports.

Open-source voice input tool supporting multiple languages and intelligent text optimization, boosting input efficiency by several times. It balances local privacy with cloud convenience, serving as a powerful assistant for productive professionals.

An open-source desktop browser based on the Firefox engine, featuring vertical tabs, workspaces, and split-screen views, emphasizing privacy protection and a modern browsing experience focused on efficiency and concentration.

Alibaba's open-source AI infographic engine uses declarative syntax + 197+ templates to generate professional charts with just one line of code, suitable for all scenarios including data visualization and news illustrations.

Alibaba's open-source AI image layering editor—automatically separates layers, precisely modifies content, no need for tedious masking, delivering efficient and professional results!

NVIDIA's open-source AI model series, featuring Nano, Super, and Ultra variants, is specifically designed for intelligent agent applications, delivering high efficiency and precision.

The world's first large model of mathematical reasoning in open source form to reach the gold medal level of the International Mathematical Olympiad (IMO), realizing the rigor of reasoning and the ability to solve difficult mathematical problems through a self-verification framework.

OceanBase is the world's first open source AI-native database, which focuses on multimodal hybrid search, minimal development and extreme security, redefining the way data and AI converge and helping developers build high-performance intelligent applications with a single click.

Meta open source revolutionary single-image 3D generation model, support one-click from 2D photos to generate high-fidelity, interactive 3D models, covering the object/human body scene, empowering e-commerce, AR/VR, film and television, and other multi-industry cost reduction and efficiency.

Ali open source SmartResume is a high-precision resume parsing system based on OCR and lightweight large models, which can convert 12 formats of resumes such as PDF/pictures into structured data in seconds, with an accuracy rate of 93.1%.

Baidu's lightweight multimodal document parsing model, with 0.9B parameters, achieves accurate recognition and structured output of complex documents in 109 languages, with world-leading performance.