DeepSeek, the Chinese startup that has gained attention with its R1 reasoning AI model, is restricting access to its API due to limited server capacity, Bloomberg reports. The company, which has achieved high performance with fewer resources than competitors, has temporarily paused customers’ ability to top up API credits to prevent service disruptions. However, existing stored balances remain unaffected. “The current top-up amount may continue to be used, so please understand!” the company stated.
Pricing Adjustments and Service Expansion
DeepSeek’s competitive pricing has raised concerns among rivals, and the company has now announced that discounts on its AI model access will end on February 8. After this date, the AI chatbot will cost 2 yuan ($0.27) per million incoming tokens and 8 yuan ($1.10) per million outgoing tokens. Once the reasoning model becomes available, the rates will increase to 4 yuan ($0.55) per million incoming tokens and 16 yuan ($2.19) per million outgoing tokens.
Since late January, DeepSeek’s services have been running at increased capacity due to the launch of its AI Assistant chatbot, reminds NIXSOLUTIONS. The company claims this AI can compete with OpenAI’s ChatGPT while being developed at a fraction of the cost. This claim has drawn scrutiny in the United States, where officials are investigating whether DeepSeek bypassed semiconductor export restrictions by acquiring Nvidia’s AI accelerators through a third party in Singapore.
We’ll keep you updated as more details emerge.