Chinese AI lab DeepSeek has announced the release of its DeepSeek-R1-Lite-Preview model, which it claims rivals OpenAI’s o1 model. The new model offers a unique feature: transparency in its reasoning process, allowing users to see its step-by-step problem-solving methods. The announcement comes two months after the launch of OpenAI. o1-Preview The model highlights the growing competition in the AI ​​reasoning space.

DeepSeek launches reasoning model to compete with OpenAI.

DeepSeek-R1-Lite-Preview can be accessed via a web chatbot, Deepsec Chatwhere users can interact with the model, limited to 50 messages per day. While detailed benchmarks and model cards have yet to be released, preliminary evaluations show that the reasoning model exhibits comparable performance to OpenAI’s benchmarks on AIME and MATH tasks. DeepSeek claims that it achieves a state-of-the-art accuracy of 91.6% on the MATH benchmark.

The introduction of DeepSeek-R1 comes as the traditional laws of scaling in AI, which suggest that increasing data and computational power will improve performance, begin to show diminishing returns. In response, companies are exploring new approaches, such as underlying reasoning models such as DeepSeek-R1. Unlike traditional models, reasoning models extend their computational processing during inference to enhance decision-making capabilities.

Despite its promising features, the new model also adheres to strict censorship protocols common to Chinese AI technology. Observations have confirmed that DeepSeek-R1 avoids sensitive political topics, such as questions about Xi Jinping or Taiwan. Users have reported successful attempts to bypass these restrictions, allowing the model to serve unfiltered content in certain situations. This aspect raises ongoing questions about the balance between functionality and regulatory compliance for AI models developed in regions with strict government oversight.

DeepSec claims that its DeepSec-R1 model — or more specifically, DeepSec-R1-Lite-Preview — matches OpenAI’s o1-Preview model on two prominent AI benchmarks, AIME and MATH. AIME evaluates model performance using other AI models, while MATH tests problem solving with a collection of word problems. However, the model is lacking. Some users on X pointed out That DeepSeek-R1, like o1, faces challenges with tic-tac-toe and other logic-based tasks.

Looking ahead, DeepSeek plans to continue its commitment to the open source AI community, releasing open source versions of its R1 models and expanding accessibility through APIs. The company is backed by High Flyer Capital Management, which is pursuing a strategy of integrating AI into trading decisions. HighFlyer’s operations include substantial investment in hardware infrastructure, boosting clusters of Nvidia A100 GPUs for model training.

Featured Image Credits: Deep Sec