Voila is an open-source, end-to-end speech model designed specifically for voice interaction. It features high-fidelity, low-latency real-time streaming audio processing capabilities, directly processing voice input and generating voice output, providing users with a smooth and intuitive experience...
kimi-thinking-preview is a multimodal thinking model developed by Dark Side of the Moon. It possesses deep reasoning capabilities and excels at solving complex problems, such as coding, mathematical, and work-related challenges. The model is based on the `reasoning_content` field...
UniTok is a unified visual word segmenter developed by ByteDance in collaboration with the University of Hong Kong and Huazhong University of Science and Technology. It supports both visual generation and understanding tasks. Based on multi-codebook quantization technology, it segments visual features into multiple small blocks, each using...
D-DiT (Dual Diffusion Transformer) is a multimodal diffusion model developed by Carnegie Mellon University, Yale University, and ByteDance Seed Labs. It unifies image generation and understanding tasks. The model combines continuous image diffusion...
NoteLLM is a multimodal large-scale language model framework launched by Xiaohongshu for note recommendation. Based on compressed embeddings of generated notes and automatic tag category generation, NoteLLM leverages the powerful semantic understanding of large-scale language models (LLMs)...
Granite 4.0 Tiny Preview is a preview version of the smallest model in IBM's Granite 4.0 language model family. Granite 4.0 Tiny Preview offers extremely high computational efficiency and a compact model structure for...
Gemini 2.5 Pro (I/O version) is an upgraded version of the Gemini 2.5 Pro multimodal AI model released by Google, specifically version number Gemini 2.5 Pro Preview 05-06. The model has achieved a significant breakthrough in programming capabilities...
VPP (Video Prediction Policy) is the first large-scale AIGC robot model launched by Tsinghua University and Xingdong Era. Based on a pre-trained video diffusion model, it learns from a large amount of video data on the Internet to directly predict future scenes...
Amazon Nova Premier is Amazon's most powerful multimodal AI model, capable of handling text, image, and video input (excluding audio). It excels at handling tasks requiring deep contextual understanding, multi-step planning, and cross-platform processing...
ICEdit (In-Context Edit) is an instruction-based image editing framework developed by Zhejiang University and Harvard University. It leverages the powerful generation capabilities and context-awareness of the Massive Diffusion Transformer...
FlexiAct is a novel action transfer model jointly developed by Tsinghua University and Tencent ARC Labs. Given a target image, FlexiAct can transfer actions from a reference video to the target subject, even in situations with significant spatial structural differences...
Klavis AI is an open-source MCP (Multimodal Communication Protocol) integration platform that helps AI applications quickly connect to production-grade MCP servers and clients. The platform provides...
Oli is an open-source intelligent code assistant that provides robust programming support for developers. It's based on a modern hybrid architecture, combining the high performance of a Rust backend with the interactive terminal interface of a React/Ink frontend. Oli supports multiple...
Mistral Medium 3 is a multimodal language model developed by Mistral AI. The model achieves a balance between performance and cost, approaching or even reaching the performance level of the Claude Sonnet 3.7 model, while costing only 1/8 of the latter (per...
Open Computer Agent is a free cloud-based AI agent tool from Hugging Face. It supports running on Linux virtual machines and performs user-specified tasks based on pre-installed programs (such as Firefox), for example...
Absolute Zero is a novel language model inference training method developed by the Tsinghua University LeapLab team in collaboration with the Beijing General Artificial Intelligence Research Institute (NLCo) and Pennsylvania State University. Absolute Zero is based on a model...
Avatar IV is HeyGen's latest AI digital human model, used to quickly create realistic videos. Users only need to upload a photo and a script or audio clip to generate natural and smooth videos. The model is based on audio-driven facial expressions...
Insert Anything is a context-editing-based image insertion framework jointly developed by researchers from Zhejiang University, Harvard University, and Nanyang Technological University. The framework seamlessly inserts objects from a reference image into a target scene...
QLIP (Quantized Language-Image Pretraining) is a visual tokenization method introduced by NVIDIA and others, combining high-quality image reconstruction and zero-shot image understanding capabilities. QLIP's binary spherical quantization (BSQ)...
ZeroSearch is an innovative large-scale model search engine framework open-sourced by Alibaba's Tongyi Lab. It leverages reinforcement learning to incentivize the search capabilities of large models without interacting with real search engines. The framework relies on pre-trained knowledge from large models...
Chatlog is an open-source chat log analysis tool that supports efficient parsing, visualization, and mining of chat logs from mainstream chat tools such as WeChat, QQ, and Telegram. The tool supports retrieving data from local database files and provides intelligent analysis...
SuperEdit is a command-guided image editing method jointly developed by ByteDance's Intelligent Creation Team and the Computer Vision Research Center at the University of Central Florida. It improves the accuracy and effectiveness of image editing by optimizing supervisory signals. SuperEdit...
FlowGram is an open-source, node-based, visual workflow building engine developed by ByteDance. It helps developers quickly build workflows with fixed or freely connected layouts. It supports two layout modes: fixed layout is suitable for sequential workflows...
Mogao is a fully-fledged interleaved multimodal generative model developed by ByteDance's Seed team. Its architecture employs dual visual encoders, combining a variational autoencoder (VAE) and a visual transformer (ViT), enabling better visual understanding...