DeepSeek restricts access to models as demand exceeds server capacity

DeepSeek AI is restricting access to its application programming due to an overwhelming response from the market which is constraining its capacity.

The Chinese start-up rattled markets across the globe with the release of it models in January which resulted in investors selling off tech stock due to fears that the emergence of low-cost Chinese AI model would threaten the dominance of stocks like Nvidia, which saw billions of dollars wiped off that week.

DeepSeek disabled the top-up button

Now the Chinese AI startup has revealed it will temporarily halt and restrict access to its application programming interface service due to limited server capacity. The company on Thursday disabled the top-up button on its platform website. According to SCMP, DeepSeek platform allows users to engage with its application programming interfaces or APIs. These serve as a bridge that allows developers and apps to interact with the company’s AI models.

In a post on its website, DeepSeek indicated it had suspended customers’ ability to top up their API credits to avoid any broader impact on the services. The company, however, explained that existing stored values would not be affected.

“The existing recharge amount can continue to be used, please understand!” the company said.

Since the launch of its chatbot, the Chinese AI startup received an overwhelming response from the market. As a result, its services have also been overwhelmed with demand since January after the firm debuted its AI chatbot which it claimed was developed at a fraction of the cost of other products.

The AI startup also claimed the superiority of its model in a technical report adding that its model outclassed rivals from OpenAI and Stability AI on rankings for image generation using text prompts.

According to Bloomberg, DeepSeek previously restricted signups for new users to people with new mainland China telephone numbers.

DeepSeek new models were eye-opening for US firms

The excitement around DeepSeek’s AI model in January sparked a $1 trillion rout in US European tech stocks. However, this was also an eye-opener as it ignited urgent efforts by the US to close gaps in its current export restrictions on chips that are used for AI applications.

The US government launched a probe into DeepSeek’s AI model to investigate data protection and data security concerns while in Ireland, its Data Protection Commission asked for further details on the startup’s operations.

Specifically, the AI startup is under probe by American officials who want to establish if it acquired Nvidia semiconductors through third parties in Singapore to dodge the restrictions.

Other regulators in South Korea and France have also started investigating how DeepSeek handles user data, which it stores in servers in China.

Now, DeepSeek, which also shocked the market and some competitors with its low pricing revealed in its posting that its discounts for access to its models would end on Feb. 8. The company has put a price tag of 2 yuan per million input and 8 yuan per million output tokens.

According to Bloomberg, when its reasoning model goes online, DeepSeek will charge 4 yuan per million input tokens and 16 yuan per million output tokens.

Source