Software size: 9.16 2025-10-23

-V3 full-blood version software is a high-performance AI model that has made a major breakthrough in the field of natural processing. It is officially released by Qingyun Technology's Computing Cloud Cornerstone Intelligence. DeepSeek-V3 is a large language model 1 released by Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co., Ltd. on December 26, 2024. It is based on a self-developed hybrid...

Software size: 9.16 2025-10-23

-R1 is a large open source reasoning model launched by China Deep Search Company. It is based on reinforcement learning technology and specializes in complex mathematics, code and natural language reasoning tasks. It takes 5 times the response speed of GPT-4 and only one-tenth the cost of OpenAIo1 as its core advantages. It supports local deployment (Ollama framework) and cloud invocation, empowering...

Software size: 10.99 2025-10-23

VL is an open source visual-language model series developed by Hangzhou Deepin Search Artificial Intelligence Basic Technology Research Co., Ltd. It uses a hybrid visual encoder and can process high-resolution images, showing excellent performance in visual-language benchmark tests. For example, DeepSeek-VL2 encodes dynamic high-resolution vision...

Software size: 37.6 2025-10-23

The Mac version is an AI smart assistant application launched by Hangzhou Deepin Search Company. It is an AI smart assistant adapted to the Apple system. It provides web-based and locally deployable usage forms. The interface is in line with macOS operating habits. The core supports long document parsing, multi-language creation, code generation and debugging, and can also be implemented offline through local deployment...

Software size: 9.16 2025-06-05

Math is a mathematical reasoning model launched by DeepSeek, focusing on solving complex problems. It is based on a 7B parameter architecture, trained through a corpus of 35.5 million mathematics and 120 billion Tokens, and supports algebra, calculus, optimization problems and mathematical theorem proofs. In the MATH benchmark test, its accuracy reached 51.7%...

Software size: 9.16 2025-06-05

LLM is an open source large-scale model (LLM) launched by related technology companies (such as the version independently developed by Philosophy Park, etc.). It is driven by long-termism and is committed to promoting the development of language models. It uses only models based on the Transformer architecture and can effectively support a variety of AI applications, such as text generation, code completion and complex data analysis...

Software size: 9.16 2025-06-05

Coder is a series of generative models developed by the DeepSeek team, focusing on improving efficiency through large-scale training and intelligent algorithms. It is based on mixed training of 2 trillion Tokens (87% code + 13% Chinese and English natural language), supports model versions from 1B to 33B parameters, and covers development needs of different scales1. This model features...