Dilemma of Free AI Models: DeepSeek Limits Conversations Due to Computing Shortages

Chinese artificial intelligence (AI) startup DeepSeek has recently imposed restrictions on long conversations and daily usage due to a shortage of computing resources. This has raised questions about the sustainability of free service models in China's AI industry.


According to a report by China’s Kechuangban Daily on November 16, DeepSeek has implemented usage measures for long conversations on its website and mobile app.


Several users reported receiving notifications from DeepSeek stating, 'Long conversations that have been inactive for more than seven days are no longer supported due to a lack of computing resources.'


The character limit for input has been reduced to approximately 40,000 characters. Some users indicated that conversations inactive for over 72 hours are also unsupported.


Daily usage limits have also been enforced. Users reaching their daily limit receive a message stating, 'You have reached your usage limit due to server congestion.'


This is not the first service disruption for DeepSeek. On November 7, some services on the DeepSeek website and app programming interface (API) were interrupted, displaying a message that read, 'The server is congested; please try again later.' Earlier, on October 15, there was a service outage lasting about five hours that affected the website, app, and API services.


According to Yonhap News, unlike previous service interruptions, this time DeepSeek cited a lack of computing resources as the direct reason for limiting user access.


DeepSeek is a prominent company that has sparked a price-cutting competition in China's AI industry by offering high-performance AI models at low costs. Chinese AI companies have rapidly gained market share by promoting free or low-cost services to attract users.


Market research firm QuestMobile estimates that DeepSeek had approximately 123 million monthly active users in China as of September, making it the third-largest AI app in the country. Despite its vast user base, analysts warn that as the number of users increases, inference costs may rise, putting pressure on the sustainability of free and low-cost services.





* This article has been translated by AI.