Global AI benchmark rankings such as LMarena

Chinese DeepSeek Unveils 27x Cheaper Open-Source Model Rivaling OpenAI’s o1

Jan 24, 2025, 01:47 PM

Chinese AI startup DeepSeek has released an open-source reasoning AI model called DeepSeek-R1 under the MIT license. The model, which has 70 billion parameters, is said to rival OpenAI's o1 in performance, matching or exceeding it on multiple benchmarks, particularly in technical domains such as coding and mathematics. DeepSeek-R1 was developed in just two months at a cost of $5.5 million, significantly less than the billions reportedly spent by US companies like OpenAI and Meta. The model is 27 times cheaper than OpenAI's o1 and is ranked in the top three in LMarena rankings. The model utilizes fewer high-end chips, demonstrating the company's focus on efficiency and resource optimization in the face of US export controls on advanced semiconductors. In response to DeepSeek's advancements, OpenAI CEO Sam Altman announced that the free tier of ChatGPT will receive an upgrade to O3-mini. The emergence of DeepSeek-R1 highlights the rapid progress of Chinese AI startups in the global AI race and raises questions about the competitive landscape and future of AI technology development.