RustGPT is a Transformer architecture language model written in Rust. Built from scratch, RustGPT does not rely on any external machine learning frameworks and uses only ndarrays for matrix operations.
Tongyi DeepResearch is an open-source deep research agent launched by Alibaba, designed specifically for long-term, deep information retrieval tasks. It boasts 30 billion parameters, activating 3 billion parameters per session, and supports ReAct mode and deep learning...
WebResearcher is an iterative deep research agent launched by Alibaba's Tongyi Labs, belonging to the Tongyi DeepResearch family. Based on an innovative iterative deep research paradigm, it simulates the cognitive workflow of human experts and can...
WebWeaver is a new dual-agent framework developed by Alibaba's Tongyi team, belonging to the Tongyi DeepResearch family, and is used for open-ended deep research. WebWeaver simulates the human research process, dividing tasks into planning (exploration...
MCP Registry is a centralized platform launched by GitHub that provides developers with a centralized way to discover and install MCP servers. The platform solves the problem of MCP servers being scattered across multiple registries, random repositories, and community discussions...
Octofriend is an intelligent programming assistant that can switch freely between large language models (LLMs) such as GPT-5 and Claude, just like a close companion, and select the optimal model according to the task requirements.
VoxCPM is a 0.5B parameter speech generation model jointly developed by Wallfacer and Tsinghua University Shenzhen International Graduate School. It has achieved industry-leading levels in speech synthesis naturalness, timbre similarity, and prosodic expressiveness. VoxC...
InternVLA-A1 is a large-scale embodied manipulation model jointly released by the Shanghai Artificial Intelligence Laboratory and the National-Local Jointly Constructed Humanoid Robot Innovation Center. It possesses integrated capabilities of understanding, imagination, and execution, enabling it to accurately complete tasks. The model...
Ling-V2 is a family of large-scale language models based on the MoE architecture launched by the Ant Financial team. The first version, Ling-mini-2.0, has 16 billion total parameters, with only 1.4 billion parameters activated for each input token.
FastMTP is Tencent's self-developed Large Language Model (LLM) inference acceleration technology. It optimizes Multi-Label Prediction (MTP) technology, replacing traditional multiple independent modules with a single MTP head sharing weights, and combines language-aware lexical compression and self-regulation...
Xiaomi-MiMo-Audio is Xiaomi's first native end-to-end large-scale speech model, open-sourced by Xiaomi. Based on an innovative pre-training architecture and hundreds of millions of hours of training data, the model is the first in the speech domain to achieve In-Context Learning (ICL)...
Lucy Edit Dev is an open-source, text-based video editing model developed by the Decart AI team. It can perform various editing operations on videos based on simple text prompts, such as changing clothing, replacing characters, inserting objects, and transitioning scenes...
Wan2.2-Animate is a motion generation model developed by the Alitongyi team. The model supports both motion imitation and role-playing modes, accurately replicating facial expressions and movements from a performer's video to generate highly realistic...
Lego-Edit is an open-source, instruction-based image editing framework from Xiaomi. It leverages the generalization capabilities of a Multimodal Large Language Model (MLLM) to enable flexible image editing. It employs a model-level toolkit, including various efficiently trained models...
Codexia is a powerful cross-platform AI coding desktop application that supports a graphical interface and toolset for the OpenAI Codex CLI, enhancing the development experience.
Kronos is the first candlestick chart model for financial markets, jointly open-sourced by Tsinghua University and Microsoft Research Asia. It analyzes candlestick data for assets such as stocks and cryptocurrencies, including opening price, highest price, lowest price, etc.
OneSearch is an end-to-end generative framework for e-commerce search launched by Kuaishou. It optimizes the cascading architecture of traditional e-commerce search, improving search accuracy and user experience. Three major innovations include: Keyword Enhancement Hierarchical Quantization Coding (KHQ)...
LSP (Language Self-Play) is a reinforcement learning method proposed by Meta to address the dependence of large language models on large amounts of high-quality training data. The core idea of LSP is to utilize a self-play mechanism, allowing models with the same model to...
TrafficVLM is a traffic visual language model launched by AutoNavi based on large-scale model technology. Through its traffic twin reconstruction capabilities, it transforms massive amounts of real-time traffic data into dynamic twin video streams, constructing a "digital..." synchronized with the real world.
DeepSeek-R1-Safe is a large-scale security model derived from DeepSeek, jointly developed by the School of Cyberspace Security at Zhejiang University and Huawei. Based on Huawei's Ascend chip and the MindSpeedLLM framework, the model constructs a security corpus...
Granite-Docling-258M is a lightweight visual language model from IBM, designed for efficient document conversion. The model can convert documents into machine-readable formats while fully preserving elements such as layout, tables, and formulas.
LongCat-Flash-Thinking is a large-scale inference model with 560 billion parameters launched by the Meituan team. Based on a hybrid expert (MoE) architecture, it supports dynamically activating 18.6 billion to 31.3 billion parameters according to demand, balancing computational...
LatticeWorld is a software developed by NetEase, City University of Hong Kong, Beijing University of Aeronautics and Astronautics, Tsinghua University, and other institutions. It combines large language models with the industrial-grade 3D rendering engine Unreal Engine 5 (UE5)...
DeepSeek-V3.1-Terminus is an AI language model developed by the DeepSeek team. It's an upgraded version of DeepSeek-V3.1, primarily optimizing language consistency and agent capabilities, and addressing issues related to mixed Chinese and English text and abnormal characters...