Alibaba said Thursday that the new model, named 'Qwen2.5-Omni-7B', is a multimodal model, so it can process text, images, audio and videos, while generating real-time text and natural speech responses.
Notably, it is relatively small at 7 billion parameters -- a measure of an AI model's size and complexity -- meaning it can potentially run on a smartphone without needing a constant connection to a server. That could make it a perfect fit to equip Apple's iPhones with AI capabilities in China.
"This sets a new standard for optimal deployable multimodal AI for edge devices like mobile phones and laptops," Alibaba said in a statement.
Apple is awaiting regulatory approval for a partnership with Alibaba to be able to offer Apple Intelligence -- its AI offering -- on devices purchased in mainland China.
Apple needs a boost as iPhone shipments in China fell 17% to 42.9 million in 2024, according to research firm Canalys. A planned AI upgrade for its Siri virtual assistant outside of China has also been delayed, prompting analysts to cut their forecasts for iPhone sales growth this year.
Meanwhile, Alibaba needs to prove that it can outcompete local Chinese AI rivals such as DeepSeek while also being able to rival Western companies despite criticizing the scale of their spending. Alibaba said recently that it would invest at least 380 billion yuan ($52.4 billion) in cloud computing and AI over the next three years. By comparison, Amazon.com plans capital expenditure of $105 billion this year alone.
Alibaba Chairman Joe Tsai said he was "astounded" by U.S. technology companies' on Tuesday at the HSBC Global Investment Summit in Hong Kong and questioned the need for spending hundreds of billions of dollars on AI data centers.
Alibaba was a Barron's stock pick for 2025, noting the potential for its low valuation to increase if investors warmed to the Chinese market.
Comments