The AI company DeepSeek, backed by a large Chinese investment fund, has just announced a new AI model designed to compete directly with OpenAI's o1.
DeepSeek-R1 is an advanced AI model that uses a reasoning method, allowing it to verify the accuracy of answers by deeply analyzing and taking extra time to think before responding.
Similar to OpenAI's o1 model, DeepSeek-R1 performs sequential actions to reach a result, which can take several seconds depending on the complexity of the question.
In tests, DeepSeek-R1 achieved performance equivalent to OpenAI's o1-preview on two popular evaluation metrics: AIME and MATH. However, DeepSeek-R1 is not perfect; some users found that the model struggled with simple problems and could be "hacked" to provide unsafe responses, including sensitive content.
DeepSeek has implemented strict censorship measures to avoid sensitive political topics. This reflects regulations from the Chinese government, requiring AI models to demonstrate "core socialist values." Additionally, these restrictions may be enforced through a blacklist of data sources not allowed for training.
The launch of DeepSeek-R1 comes amidst traditional AI models facing limits of "scaling laws" - the theory that increasing data and computational power will improve AI capabilities. Instead, companies are exploring new directions like "test-time compute", a technology used in DeepSeek-R1, allowing the model to utilize additional processing time when executing tasks.
DeepSeek impresses not only with its technology but also with its massive investment. Funded by High-Flyer Capital Management - a quantum investment fund in China - DeepSeek has a server cluster containing 10,000 Nvidia A100 GPUs, costing up to 138 million USD. The company also announced it would open-source DeepSeek-R1 and provide an API for the community.
DeepSeek has previously pressured domestic competitors like ByteDance, Baidu, and Alibaba, forcing them to lower prices or offer some AI models for free to maintain competitiveness. With DeepSeek-R1, the company continues to strengthen its position in the AI field, marking an important advance in the global technology race.