DeepSeek's New AI Model Achieves Gold Medal Status in Competitions
Full Transcript
DeepSeek, a Chinese startup, recently announced that its new AI model, DeepSeekMath-V2, achieved gold medal status at two prestigious mathematical competitions: the International Mathematical Olympiad 2025 and the Chinese Mathematical Olympiad 2024.
This significant milestone was reported by Matthias Bastian from The Decoder on November 29, 2025. The DeepSeekMath-V2 model showcased exceptional performance, solving five out of six problems at the IMO 2025 and obtaining an impressive score of 118 out of 120 on the Putnam 2024 competition.
This model is notable not only for its achievements but also for its innovative architecture that employs a generator-verifier approach, which enhances the model's capability for self-verifying mathematical reasoning.
The model's architecture includes a verifier specifically trained to score proofs and a generator that can identify and correct its own mistakes, a process described as a self-checking loop. The implications of this achievement extend beyond academia, as it positions DeepSeek alongside major players like OpenAI and Google in the realm of advanced AI technologies.
DeepSeek has made the weights of the DeepSeekMath-V2 model publicly available under the Apache 2.0 license on platforms like Hugging Face and GitHub, making it the first open-source model capable of winning gold in the IMO.
This release is part of a broader trend in AI where open-source models are gaining traction, allowing developers and researchers globally to leverage cutting-edge technology without the barriers typically associated with proprietary software.
Experts on social media have highlighted that DeepSeek's achievement represents a significant leap in mathematical reasoning capabilities within AI, indicating that the model might reshape how machines approach complex problem-solving in the future.
The broader developer community is encouraged by this development, emphasizing the importance of open-source contributions in advancing the field of AI. Overall, DeepSeek's DeepSeekMath-V2 not only marks a technological milestone but also serves as a testament to the potential of AI in tackling complex challenges in mathematics and beyond.