Alibaba Cloud, the cloud technology arm of Alibaba Group, unveiled two Large Language Models (LLMs): Qwen-72B and Qwen-1.8B. These models, with 72 billion and 1.8 billion parameters respectively, are versions of the company’s proprietary foundation model Tongyi Qianwen.
Alibaba Cloud’s Qwen-72B and Qwen-1.8B are now available on ModelScope and Hugging Face, Alibaba’s collaborative artificial intelligence (AI) platform.
“Building up an open-source ecosystem is critical to promoting the development of LLM and AI applications building,” said Jingren Zhou, CTO of Alibaba Cloud. “We aspire to become the most open cloud and make generative AI capabilities accessible to everyone. To achieve that goal, we will continue to share our cutting-edge technology and facilitate the development of the open-source community together with our partners.”
Alibaba Cloud launches upgraded version of its AI model
Alibaba Cloud rolls out open-source vision language model
Audio LLMs
Expanding beyond text-based models, Alibaba Cloud has introduced multimodal LLMs like Qwen-Audio and Qwen-Audio-Chat, catering to audio understanding for both research and commercial purposes. Their parameter range spans from 1.8B to 72B, incorporating audio and visual features.
The flagship 72-billion-parameter model, trained on over 3 trillion tokens, outperforms other open-source models across ten benchmarks, showcasing superior multitask accuracy, code generation, and arithmetic problem-solving abilities. Notably, it excels in role-playing and language style transfer, ideal for personalized chatbot applications.
The model also exhibits proficiency in tackling a variety of intricate tasks, including role-playing and language style transfer, referring to the ability of the LLM to assume a specific role or persona and generate more contextually relevant responses consistent with the persona. Such features can be useful in AI applications such as personalized chatbots.
Multi-modal LLMs
Alibaba Cloud offers free access to the Qwen-72B model for research purposes and commercial use for companies with fewer than 100 million monthly active users. The company also open-sourced a lightweight 1.8-billion-parameter model for edge devices facilitating deployment on resource-constrained devices like phones.
This initiative aligns with Alibaba Cloud’s commitment to multi-modal LLMs capable of understanding various data types, building on their previous releases like Qwen-VL and Qwen-VL-Chat, which cater to visual information processing.
Since August, these open-source LLMs have amassed over 1.5 million downloads on ModelScope and Hugging Face, fostering an expansive developer community exceeding 2.8 million active members and over 100 million model downloads.