PProfiling Data is performance analysis data from DeepSeek's open-source training and inference framework. It is based on detailed information captured during program execution by PyTorch Profiler and is used to analyze and optimize software performance.
Phi-4-Multimodal is Microsoft's latest multimodal language model, boasting 5.6 billion parameters and integrating speech, vision, and text processing into a unified architecture. The model performs exceptionally well on multiple benchmarks, including automatic...
R1-Onevision is an open-source multimodal language model focused on complex visual reasoning tasks. Based on a fine-tuned version of Qwen2.5-VL, it integrates visual and textual data to accurately interpret multimodal information. In mathematics...
Phi-4-Mini is the latest addition to Microsoft's Phi-4 series of small language models focused on text tasks, with 3.8 billion parameters. It is based on a dense decoder-only Transformer architecture, combined with grouped query attention (G...
IndexTTS is an industrial-grade, controllable text-to-speech (TTS) system launched by Bilibili. Developed based on the XTTS and Tortoise models and incorporating GPT-style generation techniques, it efficiently converts text into natural and fluent speech...
Scribe, developed by ElevenLabs, is a high-precision speech-to-text model designed for multilingual and complex audio environments. It supports 99 languages, achieving transcription accuracy of 96.7% for English and 98.7% for Italian. In less commonly spoken languages...
VideoGrain is a zero-shot, multi-granularity video editing framework developed by the University of Technology Sydney and Zhejiang University. It enables fine-grained video modification at the category, instance, and component levels. VideoGrain is based on adjusting spatiotemporal cross-attention and self...
Tencent's Hunyuan Turbo S is a new generation of fast-thinking model launched by Tencent. The model adopts an innovative Hybrid-Mamba-Transformer fusion architecture, effectively reducing the computational complexity of traditional Transformers and minimizing KV-Cache...
GPT-4.5 is OpenAI's latest large-scale language model, currently the largest and most powerful chatbot model. Based on extended unsupervised learning, it improves pattern recognition, knowledge breadth, and creative generation capabilities, while reducing illusions...
3FS (Fire-Flyer File System) is a high-performance distributed file system from DeepSeek, designed specifically for AI training and inference tasks. 3FS uses modern SSDs and RDMA network technology, aggregating thousands of SSDs based on a discrete architecture...
CorrDiff is a generative AI model from NVIDIA used to downsample low-resolution global weather data to high-resolution data, improving the accuracy and efficiency of weather forecasting. It employs a two-step data processing method: first, through...
Smallpond is a lightweight data processing framework developed by DeepSeek, built on DuckDB and 3FS, and designed for high-performance and large-scale data processing. Smallpond supports processing petabyte-scale datasets, leveraging DuckDB's...
SongGen is a single-stage autoregressive Transformer model developed by Shanghai AI Lab, Beijing University of Aeronautics and Astronautics, and the Chinese University of Hong Kong for generating songs from text. SongGen is based on lyrics and descriptive text (such as instruments, style, etc.)...
VideoFusion is an open-source short video splicing and processing software designed for efficient video editing. It supports automatic removal of black borders, watermarks, and subtitles from videos, and can automatically rotate videos to landscape or portrait orientations to adapt to different playback scenarios...
Baichuan-Audio is an end-to-end audio language model developed by Baichuan Intelligence. It supports seamless integration of audio understanding and generation functions, enabling high-quality, controllable real-time Chinese-English bilingual dialogue. Baichuan-Audio is based on multi-code...
AIMv2 is an open-source multimodal autoregressive pre-trained visual model from Apple that improves the performance of visual models through deep fusion of images and text. It employs an innovative pre-training framework that divides images into non-overlapping image patches and integrates text...
VidSketch is an innovative video generation framework developed by the State Key Laboratory of CAD & CG and the School of Software at Zhejiang University. It generates high-quality video animations based on hand-drawn sketches and simple text prompts. VidSketch is based on a hierarchical...
HippoRAG 2, developed by Ohio State University, is a retrieval-enhanced generative (RAG) framework that addresses the limitations of existing RAG systems in simulating the dynamics and associative nature of human long-term memory. HippoRAG 2 is based on personalized PageRank...
LDGen is an innovative text-to-image synthesis technology that improves the quality and semantic consistency of text-to-image generation by combining Large Language Models (LLMs) with diffusion models. Through hierarchical title optimization and human-instruction techniques,...
AI-Infra-Guard is an open-source, efficient, lightweight, and easy-to-use security assessment tool for AI infrastructure developed by Tencent. It can discover and detect potential security risks in AI systems. AI-Infra-Guard supports 28 AI frameworks...
Mahilo is a flexible multi-agent framework that supports the creation of multi-agent systems that interact with humans. Mahilo supports real-time voice and text communication, allowing agents to autonomously share context and information while maintaining human oversight of the interactions...
WorldCraft, developed by the Hong Kong University of Science and Technology, is a 3D world creation and customization system based on Large Language Model (LLM) agents. It's used to create and customize realistic 3D virtual worlds. Based on natural language interaction, it allows users to easily generate...
Mobius is an advanced seamless looping video generation technology that uses artificial intelligence algorithms to generate infinitely looping video content from text descriptions. At its core lies a powerful AI model that automatically identifies key elements in the video and generates smooth...
WarriorCoder is a Large Language Model (LLM) for code generation, developed by the School of Computer Science and Engineering at South China University of Technology and Microsoft. It generates high-quality training data by simulating adversarial interactions between expert models, thereby improving model performance. ...