LazyLLM is an open-source low-code platform that helps developers quickly and cost-effectively build multi-agent large language model applications. Through a streamlined development process, it supports one-click deployment and cross-platform operation, reducing the complexity of AI application development...
Memobase is an open-source, user-profile-based long-term memory system designed for generative artificial intelligence (GenAI) applications. Memobase provides user memory capabilities for applications such as virtual assistants and educational tools, helping AI remember...
MeteoRA is a multi-task embedding framework for large language models (LLMs) developed by a research team in the Department of Computer Science and Technology at Nanjing University. It integrates multiple task-specific LoRAs (low-rank adapters) into a single base model...
MME-CoT is a technology jointly developed by the Chinese University of Hong Kong, Shenzhen, ByteDance, Nanjing University, Shanghai Artificial Intelligence Laboratory, University of Pennsylvania, Tsinghua University, and other institutions for evaluating large-scale multimodal models...
Claude 3.7 Sonnet, developed by Anthropic, is the world's first hybrid reasoning model, featuring both a "standard mode" and an "extended thinking mode." In standard mode, Claude 3.7 Sonnet can quickly generate...
DeepEP is DeepSeek's first open-source EP (Expert Parallel) communication library specifically designed for hybrid expert model (MoE) training and inference. It provides high-throughput and low-latency full-to-full GPU kernels, supporting intra-node...
QwQ-Max is a deep inference model developed by Alibaba based on Qwen2.5-Max, and is part of the Qwen series. QwQ-Max-Preview is being released as a preview version; QwQ-Max will soon be officially released and fully open-source. QwQ-Max...
FantasyID is a novel identity-preserving video generation (IPT2V) framework developed by Alibaba Group and Beijing University of Posts and Telecommunications. It generates high-quality, identity-consistent videos based on enhanced facial knowledge. FantasyID is based on the Diffusion Transformer (D...
SigStyle is a new signature style transfer framework developed by Jilin University, the School of Intelligent Science and Technology of Nanjing University, and Adobe. It supports transferring unique visual features (such as geometric structure, color scheme, brushstrokes, etc.) from a single style image.
VLM-R1 is a visual language model developed by Om AI Lab, based on reinforcement learning technology. It accurately locates target objects in images using natural language commands, such as finding the corresponding image region based on the description "the red cup in the picture." ...
FacePoke is an open-source, real-time facial editing tool based on AI technology. Users can easily edit head orientation (e.g., tilting head up, tilting head down, shaking head left and right) and facial expressions (e.g., eye opening and closing) in photos using simple mouse drag-and-drop operations.
BFS-Prover is an automated theorem proving system based on Large Language Models (LLM) developed by ByteDance's Doubao Big Model team. It improves upon the traditional Breadth-First Search (BFS) algorithm by incorporating expert iteration and direct preference optimization...
PySpur is an open-source, lightweight, visual AI agent workflow builder that simplifies the development process of AI systems. Its drag-and-drop interface allows users to quickly build, test, and iterate AI workflows without writing complex code. ...
Wan2.1 is an open-source AI video generation model from Alibaba Cloud, possessing powerful visual generation capabilities. Wan2.1 supports both text-based and image-based video tasks, and includes models of two sizes. The professional version with 14 bytes of parameters excels at generating complex motion...
AgentRefine is an intelligent agent synthesis framework jointly proposed by Beijing University of Posts and Telecommunications and Meituan. It improves the generalizability of agents based on large language models (LLMs) in diverse tasks through "refinement tuning"...
DeepGEMM is an open-source library from DeepSeek designed for efficient and concise FP8 matrix multiplication (GEMM), currently supporting only the NVIDIA Hopper tensor core. DeepGEMM supports GEMM operations in both normal and mixed expert (MoE) groups...
TinyR1-Preview is a 32-byte inference model jointly developed by the School of Computer Science at Peking University and 360 Security Technology Inc. Using only 5% of the parameters, the model approximates the performance of Deepseek-R1-671B. TinyR1-Preview has significant advantages in mathematical fields...
SurveyX is a system for automatically generating academic reviews based on Large Language Models (LLMs). It was jointly developed by Renmin University of China, the University of Sydney, and Northeastern University of China. Based on user-provided paper titles and keywords, it quickly generates...
Flame is an open-source, multimodal AI model that supports converting UI design screenshots into high-quality, modern front-end code. Based on visual language modeling, automated data synthesis, and a structured training process, Flame generates code that conforms to modern front-end technologies such as React...
OctoTools is an open-source intelligent agent framework developed by Stanford University. It uses scalable tools to solve complex inference tasks. OctoTools encapsulates tool functionality using standardized tool cards, eliminating the need for additional training...
PhotoDoodle is an artistic image editing framework jointly developed by the National University of Singapore, Shanghai Jiao Tong University, Beijing University of Posts and Telecommunications, ByteDance, and the Tiamat team. It learns the unique style of artists based on a small number of samples to achieve...
video-subtitle-master is a tool for batch generating subtitles for videos or audio. Developed based on the open-source project VideoSubtitleGenerator, it supports batch subtitle generation for videos or audio and translating subtitles into other languages...
DualPipe is an innovative bidirectional pipelined parallel technology open-sourced by DeepSeek, primarily used to improve the training efficiency of large-scale deep learning models. The core idea is to divide the model training process into two independent pipelines—forward computation...
EPLB (Expert Parallelism Load Balancer) is an expert parallel load balancer introduced by DeepSeek, designed to solve the problem of uneven load distribution among different expert models during large-scale model training. EPLB is based on...