Vibe Kanban is an open-source Kanban tool that helps developers efficiently manage AI coding agents, such as Claude Code, Gemini CLI, and Codex. Through a visual Kanban board, users can easily switch between and orchestrate multiple coding agents...
ClipSketch AI is an open-source, AI-driven content creation tool designed for video creators, social media operators, and content creators. It can parse video links from Bilibili and Xiaohongshu, automatically extract keyframes, and convert them...
IQuest-Coder-V1 is an advanced AI model from IQuestLab focused on code generation, understanding, and optimization. It offers various parameter scales (7B, 14B, 40B) and versions (Instruct, Thinking, Loop)...
RedInk is an open-source AI-powered content creation tool for Xiaohongshu (Little Red Book), dedicated to simplifying the content creation process. After users input a theme, the tool can quickly generate eye-catching titles, high-quality text, and beautiful images, achieving "one sentence to generate a complete set..."
mHC (Manifold-Constrained Hyper-Connections) is a novel neural network architecture design method introduced by the DeepSeek team, which solves the problems of traditional hyper-connections (HC) architecture in large-scale training...
AutoMV is an automated multi-agent system developed by M-A-P, Beijing University of Posts and Telecommunications, Nanjing University's NJU-LINK Laboratory, and other institutions. It supports converting songs into coherent, rhythmically aligned music videos. The system detects music information...
HY-Motion 1.0 (Hunyuan Motion 1.0) is an open-source, billion-parameter text-to-3D motion generation model developed by Tencent's Hunyuan team. Based on the Diffusion Transformer (DiT) architecture and stream matching mechanism, it requires only one line of code...
Tencent-HY-MT1.5 is an open-source translation model from Tencent's Hunyuan platform, comprising two versions: Tencent-HY-MT1.5-1.8B and Tencent-HY-MT1.5-7B. The model supports translation between 33 international languages and 5 minority/dialect languages, covering...
Computer Use Preview is an open-source AI-powered browser automation tool from Google. Based on the visual recognition capabilities of the Gemini model, it simulates human interaction with web pages through a "screenshot-analysis-action" process, without relying on fixed...
PersonaLive is an open-source real-time digital human live streaming tool developed by the University of Macau and GVC Labs. The tool can run on a single consumer-grade graphics card with 12GB of VRAM, enabling unlimited portrait animation generation and supporting low-latency interaction.
Yume 1.5 is an interactive world generation model developed by the Shanghai Artificial Intelligence Laboratory and Fudan University, among others. It can generate realistic, continuous, and explorable virtual worlds from single images or text prompts.
Step-DeepResearch is a 3.2 billion parameter deep research AI model launched by Step-Leap Star Technology. It can complete complex research tasks and generate professional reports in a single inference iteration. The model adopts a monolithic architecture design, utilizing planning, deep search, and inverse reasoning...
MAI-UI is a full-scale GUI intelligent agent foundation model launched by Tongyi Labs, possessing three core capabilities: user interaction, tool invocation, and edge-cloud collaboration. The framework utilizes an autonomously evolving data pipeline and large-scale online reinforcement learning technology...
Zhipu's new generation of large-scale foundational models boasts a total of 355 bytes of parameters and 32 bytes of activation parameters. The model achieves comprehensive advancements in areas such as real-world programming, long-context processing, reasoning ability, information retrieval, writing capabilities, and intelligent agent applications...
InstanceAssemble is a lightweight layout-to-image generation framework jointly developed by Xiaohongshu and Fudan University. Through an innovative "instance assembly attention" mechanism, the framework achieves precise layout generation from simple to complex, and from sparse to dense...
Zen Browser is an open-source browser based on the Firefox kernel, designed to enhance user productivity through efficient design and powerful features. Zen Browser optimizes the user experience for widescreen displays with its vertical tab layout and sidebar...
QwenLong-L1.5 is a large language model developed by Alibaba Tongyi Lab, focusing on long text reasoning. Based on the Qwen3-30B-A3B architecture, the model significantly improves long text reasoning capabilities through a systematic post-training scheme.
FrontierScience is a scientific AI capability assessment benchmark launched by OpenAI, specifically designed to test the expert-level reasoning capabilities of large models in the fields of physics, chemistry, and biology. It includes two subsets: the Olympic track (100 competition-level short questions)...
ML-Master 2.0 is an autonomous intelligent agent developed by the SciMaster team, comprised of members from the School of Artificial Intelligence at Shanghai Jiao Tong University, the Shanghai Institute for Algorithm Innovation, and DeepTech, for real-world machine learning research tasks. ML-Master 2.0...
QuantiPhy is the first benchmark developed by Fei-Fei Li's team at Stanford University to quantitatively evaluate the physical reasoning capabilities of visual-language models (VLMs). QuantiPhy uses over 3300 video-text instances to require models to perform physical reasoning based on visual...
Infographic is a next-generation infographic framework open-sourced by Alibaba's AntV front-end team. Based on the G2 visualization engine and Ant Design, the framework can quickly generate high-quality, stable infographics.
opcode is an open-source desktop graphical interface tool built for Claude Code. It supports transforming complex command-line operations into an intuitive visual experience, greatly improving the user experience.
MedASR is a speech recognition model developed by Google, focused on the medical field. Based on the Conformer architecture, it has 10⁵ million parameters. The model was pre-trained on a large amount of medical speech data (approximately 5000 hours), covering...
TurboDiffusion is a video generation acceleration framework jointly developed by Tsinghua University, Bionumerical Research, and the University of California, Berkeley. The framework utilizes SageAttention, SLA (Sparse Linear Attention), and rCM (Timestep Distillation)...