China’s most popular AI app gets facelifting with bytedances Doubao 1.5

On January 24, 2025By Littum

Tiktok parent Bytedance have launched an updated version of Doubao, China’s most popular consumer -oriented Artificial Intelligence (AI) app as the technology giant accelerates AI development despite US export restrictions on advanced chips.

The Beijing-based business introduced its closed source Multimodal Model Doubao 1.5 Pro on Wednesday, and emphasized a “resource-efficient” training approach that it said does not sacrifice performance.

“The model adopted an integrated train-inferences design from the pre-training phase to balance the best performance and most optimal inferencing costs,” Bytedance said in a statement, adding that it has designed a server cluster with flexible support of low-end chips to reduce AI infrastructure costs.

Do you have questions about the biggest topics and trends from all over the world? Get the answers with SCMP knowledgeOur new platform of curated content with explanations, often asked questions, analyzes and infographics brought to you by our award -winning team.

China’s major technology companies strive to catch up with their US counterparties while facing budgetary restrictions and limited access to advanced chips. This has pushed them to innovate in AI model efficiency and refine their products within the country’s closed market.

Bytedances Chatgpt-like Doubao-Bot has emerged as the most popular AI app for Chinese consumers. Photo: Simon Song Alt = Bytedances Chatgpt-like Doubao-Bot has emerged as the most popular AI app for Chinese consumers. Photo: Simon Song>

Benchmark tests have shown that Doubao 1.5 Pro stands out in half of 14 evaluations that assessed the model’s language understanding, mathematics and coding skills, domain knowledge, visual understanding and reasoning.

In some areas it did better than industry-leading AI systems from Microsoft-ho -supported Openai, Googleand Amazon.com-Supported anthropic. It also beat domestic rivals in some tests, including systems from the latest newly started tax Deepseek and Cloud Computing giant Alibaba Group HoldingOwner of the mail.

Bytedancer’s resource-efficient AI development strategy follows the success of Hangzhou-based Deepseek, which surprised the global technology society with its V3 model revealed last year. Deepseeks Open Source V3 was trained by using significantly fewer resources than Meta platforms‘Llama 3.1 while providing comparable performance.

Deepseek revealed in the model’s technical report that it was trained by using Nvidia’s H800 chips at a price of less than $ 6 million. In comparison, Anthropic CEO Dario Amodei estimated that Openai spent $ 100 million on training GPT-4o.

Bytedance did not reveal the cost or chips used to train its latest model, and did not immediately respond to a request for comment Wednesday.

The ability to promote AI models while handling resource restrictions plays a more important role in industry in the midst of an escalating technological rivalry between the United States and China.

Despite US restrictions, Chinese companies continue to make progress in AI development. The Doubao 1.5 Pro release came a few days after Deepseek and Beijing-based Moonshot AI launched new reasoning models this week.

This article originally appeared in South China Morning Post (SCMP)The most authoritative voice that has reported on China and Asia for more than a century. For more SCMP stories, please explore SCMP app or visit the SCMPs Facebook and Twitter pages. Copyright © 2025 South China Morning Post Publishers Ltd. All rights reserved.