DEPSEEK failed with popularity-access to AI models is now limited

The Chinese startup Deepseek, which caused a shock in Silicon Valley to its R1 model, capable of reasoning, due to the achievement of a high result using much smaller resources and means, announced the restriction of access to its API service due to lack of serveral capacities, writes Bloomberg.

Image source: Solen Feyissa/unsplash.com

The DeepSeek website reports on the suspension of customer opportunities to replenish their API loans in order to avoid negative consequences for the company’s services. All existing saved values ​​will not be affected, DEPSEEK added. “The current amount of replenishment can continue to be used, we ask for your understanding!” – said the company.

Also, Deepseek, whose extremely low tariffs alarmed some competitors, indicated in the report that discounts on access to her AI models would stop February 8. Further, access to Chat Bota based on AI will be paid from the rate of 2 yuan ($ 0.27) for 1 million incoming tokens and 8 yuan ($ 1.10) for 1 million outgoing tokens. When launching on the network, models with the ability to reasoning, tariffs will amount to 4 yuan ($ 0.55) for 1 million incoming tokens and 16 yuan ($ 2.19) for 1 million outgoing tokens.

Due to high demand, DeepSeek services have been working with an increased load from the end of January after the AI ​​Assistant chat boot was presented based on artificial intelligence, which, according to the company, is able to compete with the ChatGPT chat bout from Openai, and for its development It took much less funds than competitors. This statement caused doubt about the United States, and now American officials find out if DeepSeek has acquired advanced AIS-accelerators of NVIDIA through third parties in Singapore, bypassing export restrictions on the supply of semiconductors to China.