DeepSeek

DeepSeek is a new China-based startup that has created a Large Language Model (LLM) chatbot as capable as the current industry standard, ChatGPT. DeepSeek was founded in 2023.

The new models created were out of necessity made on a much lower budget than most U.S.-based programs. The United States government restricts the number of NVIDIA and other AI chips exported to China, meaning DeepSeek only had approximately 2,000 of NVIDIA’s H800 chips (now banned from export), as opposed to ChatGPT which originally trained with 10,000 chips. According to reports, it now uses an estimated 30,000.

DeepSeek-V3 was publicly released on December 26, 2024, and has experienced a major uptick in users over the past few weeks. It was followed by a release of their first reasoning, or “thinking” model, R1, on January 20th of this year. On January 27th the New York Times wrote an article on DeepSeek’s latest model, causing NVIDIA to plummet 17% on the stock market. Other stocks dropped heavily as well. On January 31st, OpenAI (ChatGPT’s parent company) released its o1-mini model, which is considered better than DeepSeek R1. I feel that they are strong in different areas, but I prefer ChatGPT because in my experience DeepSeek’s servers are usually full and not available for use.

—THEODORE HAVEL, THE CALIFORNIA TECH