DeepSeek R1 0528

DeepSeek-R1-0528 still uses the DeepSeek V3 Base model released in December 2024 as its foundation, but invested more computing power during the post-training process, significantly improving the model's depth of thought and reasoning capabilities. The updated R1 model has achieved the leading performance among all current domestic models across multiple benchmark evaluations such as mathematics, coding, and general logic, and its overall performance is already close to other top international models, such as o3 and Gemini-2.5-Pro.