New Milestone in AI: Open Source Model DeepSeek-Math-V2 Reaches Gold Medal Standard at the International Mathematical Olympiad
DeepSeek-AI has released a new language model, DeepSeek-Math-V2, specializing in mathematical reasoning. It is the first open-source model to achieve a performance equivalent to a gold medal at the International Mathematical Olympiad (IMO). This marks a significant turning point in the ability of open systems to tackle complex scientific problems.
The research team at DeepSeek-AI has introduced the latest version of their specialized language model, DeepSeek-Math-V2. The model has garnered attention for its ability to handle demanding mathematical tasks. According to its publishers, this is the first time an open-weights model has achieved a performance level meeting the requirements for a gold medal at the International Mathematical Olympiad (IMO).
The ability of AI models to solve complex mathematical problems at this caliber has long been considered a critical benchmark for progress in deep machine reasoning. Until now, the most significant advancements in this specific domain have been largely confined to closed systems from leading technology companies. For instance, in July 2024, Google DeepMind announced systems that achieved a silver-medal standard at the IMO, which was considered a major milestone at the time [1] and they finally got a gold medal performance earlier this year [2].
The fact that DeepSeek-Math-V2, as an open model, has reached a gold medal standard highlights the rapid development of open science in the field of AI and opens new possibilities for researchers worldwide to build upon this technology.
For those interested, the technical paper detailing the model and its implementation can be accessed on GitHub, and the model itself is available for download via Hugging Face.
- Technical Paper (GitHub): https://github.com/deepseek-ai/DeepSeek-Math-V2/blob/main/DeepSeekMath_V2.pdf
- Model Download (Hugging Face): https://huggingface.co/deepseek-ai/DeepSeek-Math-V2
References: [1] Google DeepMind. (2024, July 25). AI achieves silver-medal standard solving International Mathematical Olympiad problems. Retrieved from https://deepmind.google/discover/blog/ai-solves-imo-problems-at-silver-medal-level/
[2] Google DeepMind. (2025, 21. júlí). Advanced version of Gemini with Deep Think officially achieves gold-medal standard at the International Mathematical Olympiad. Sótt af https://deepmind.google/blog/advanced-version-of-gemini-with-deep-think-officially-achieves-gold-medal-standard-at-the-international-mathematical-olympiad/
