There China is gaining momentum on artificial intelligence. Its national flagship, Alibaba Cloud has unveiled an update to its Qwen (also known as “Tongyi Qianwen”) language model family with version 2.5. The firm justifies this by “response to the growing demand for AI solutions”. Indeed, since June 2023, Qwen LLMs have been deployed by more than 90,000 professional users via Alibaba Cloud's generative AI platform, Model Studio, assures the cloud subsidiary of the giant Alibaba.
An adoption that has taken place in sectors ranging from consumer electronics to automotive to gaming in China. In addition, the company cites the figure of 2.2 million enterprise users who also accessed AI services based on Qwen from DingTalk, its collaborative work and application development platform. Leveraging the OpenCompass LLM evaluation platform, Alibaba Cloud says Qwen2.5 demonstrates competitive results in various categories among existing similar LLMs, including OpenAI's GPT-4 Turbo.
Xiaomi relies on Qwen to offer more connected experiences
Among the many companies that have adopted Qwen is Xiaomi. The latter has integrated the models into its own AI assistant, Xiao Ai, to power functions such as image generation and understanding in its latest smartphone range. Users can also explore Alibaba Cloud's image generation model, Tongyi Wanxiang, and the large vision language model Qwen-VL, for different applications, including visual creation, photo analysis, and image generation. recipes from food images, all with the aim of improving the user experience.
Xiao AI is also integrated into Xiaomi's range of electric vehicles, including the SU7 model, to generate images on the vehicle's infotainment system through voice commands. The goal is always the same: to provide an enriched user experience in the vehicle with interactive entertainment options.
A breakthrough in the open source ecosystem
At the same time, the company continues its excursion into open source and upgrades Model Studio – its AI model and app development platform – with new AI development resources. For the record, its range of open source LLMs has models ranging from 0.5 billion to 110 billion parameters (Qwen 1.5), Alibaba Cloud says Qwen has seen over 7 million downloads on platforms such as Hugging Face and Github . The company also plans to share the 7 and 72 billion parameter variants of Qwen 2.0 in the coming months.
The revamped Model Studio now includes a range of advanced development tools, integrating LlamaIndex, an open source framework that makes it easy to enhance LLMs for developing applications such as chatbots and analytics tools. Another function appears: API Assistant. This should help simplify the development of sophisticated AI assistants through the seamless integration of Retrieval Augmented Generation (RAG) technology to develop contextual responses. Advanced features such as multi-agent coordination and memory management are notably supported.
Do you want to stay up to date on the latest news in the artificial intelligence sector? Register for free to the IA Insider newsletter.
Selected for you