Parlant is an open-source LLM (Large Language Model) agent behavior modeling engine that helps developers quickly create conversational agents that meet business requirements. It ensures that the AI agent's behavior conforms to business logic through natural language rule definitions...
rStar2-Agent is an open-source mathematical reasoning model from Microsoft with only 14 billion parameters. Trained using agent-based reinforcement learning, it achieves an accuracy of 80.6% in the AIME24 mathematical reasoning test, surpassing the 671 billion parameters of DeepSeek-R1.
MiniCPM 4.1 is a hybrid thinking model developed by Wallfacer. It employs the InfLLM v2 sparse attention architecture, requiring each word to be correlated with less than 5% of words, significantly reducing the overhead of long text processing. In 128K long texts...
AI Video Transcriber is an open-source video transcription and summarization tool that supports over 30 platforms, including YouTube and TikTok. The tool uses Faster-Whisper for high-precision speech-to-text conversion and leverages AI to optimize the text and correct spelling errors...
DeepDoc is an open-source deep research tool focused on in-depth research of local knowledge bases. The tool uses a research-based workflow to extract text from local resources (such as PDF, DOCX, JPG, TXT, etc.), segment the data stored in...
Visual Story-Writing is an innovative AI writing tool that helps users build and edit stories through a visual interface. The tool uses GPT-4o technology to automatically parse text, transforming it into a visual structure of events, characters, and actions...
Qwen3-ASR-Flash is the latest speech recognition model in the Tongyi Qianwen series. Based on the Qwen3 platform, it has been trained using massive amounts of multimodal and ASR data. The model supports 11 languages and multiple accents, and boasts high accuracy and high performance...
AntSK FileChunk is a semantically based intelligent text slicing tool specifically designed for processing PDF and Word documents. Based on advanced semantic analysis technology, it segments long documents into semantically complete and coherent fragments, avoiding the need for traditional text splitting...
XTuner V1 is a new generation of large-scale model training engine open-sourced by the Shanghai Artificial Intelligence Laboratory. Developed based on PyTorch FSDP, it systematically optimizes memory, communication, and load issues in training ultra-large models, supporting 1TB...
REFRAG is a high-efficiency decoding framework for Retrieval Enhancement Generation (RAG) tasks, developed by Meta Super Intelligence Lab. It optimizes large-scale language processing through a "Compress, Sense, Expand" process...
LMDeploy is a large-model inference deployment tool developed by the Shanghai Artificial Intelligence Laboratory. It significantly improves the inference performance of large models, supports various hardware architectures, including NVIDIA's Hopper and Ampere series GPUs, and enables...
Wenxin Big Model X1.1 is a deep thinking model launched by Baidu. Based on an iterative hybrid reinforcement learning framework, it significantly improves the capabilities of factuality, instruction compliance, and agent intelligence, increasing them by 34.8%, 12.5%, and 9.6%, respectively. In terms of weight...
SpikingBrain-1.0 is a large-scale brain-like spiking model developed by the Institute of Automation, Chinese Academy of Sciences. Based on intrinsic complexity, the model uses a novel non-Transformer architecture, overcoming the limitations of the Transformer architecture in handling ultra-high-density neural networks...
HunyuanImage 2.1 is an open-source text-to-image model launched by Tencent. It supports native 2K resolution, has powerful complex semantic understanding capabilities, and can accurately generate scene details, human expressions and actions.
K2-Think is an open-source inference model jointly developed by the Mohammed bin Zayed University for Artificial Intelligence (MBZUAI) in the UAE and G42. It boasts 32 billion parameters and excels in multiple fields, including mathematical reasoning, code generation, and scientific knowledge, particularly...
SRPO (Semantic Relative Preference Optimization) is a text-to-image generation model developed by Tencent Hunyuan. It designs the reward signal as a text-conditional signal, enabling online adjustment of the reward and reducing reliance on external factors...
Qwen3-Next is a new generation of hybrid architecture model open sourced by Alibaba Tongyi, which includes two versions: the instruction version (Qwen3-Next-80B-A3B-Instruct) and the thinking version (Qwen3-Next-80B-A3B-Thinking).
AutoDocs is an open-source code documentation automation tool developed by TrySita, helping developers efficiently generate and maintain codebase documentation. It generates accurate and... by deeply analyzing the codebase's Abstract Syntax Tree (AST) and build dependency graph.
AnyI2V is an innovative image and animation generation framework jointly developed by Fudan University, Alibaba DAMO Academy, and Lakeside Lab. The framework requires no large amount of training data and can transform static conditional images (such as meshes, point clouds, etc.) into dynamic videos...
MiniMax Music 1.5 is an advanced AI music generation model that supports music creation of up to 4 minutes in length. It can generate high-quality music works based on natural language descriptions such as style, mood, and scene input by the user.
FireRedTTS-2 is an advanced long-format streaming text-to-speech (TTS) system focused on multi-speaker dialogue generation. Employing a 12.5Hz streaming speech segmenter and a dual Transformer architecture, it achieves low latency, high fidelity, and multi-speaker dialogue generation...
HuMo is a multimodal video generation framework jointly proposed by Tsinghua University and ByteDance's Intelligent Creation Lab, focusing on human-centered video generation. It can generate high-quality, detailed videos from multiple modal inputs, including text, images, and audio...
Stand-In is a lightweight video generation framework developed by Tencent's WeChat Vision Team, focusing on generating identity-protected videos. By training only 1% of the base model parameters, the framework can generate high-fidelity, identity-consistent videos, and also possesses...
ST-Raptor is a tool for semi-structured table-based question answering. It requires only an Excel spreadsheet and a natural language question as input to produce accurate answers.