DeepSeek launches AI model achieving gold-level maths scores
An open-source AI model, Math-V2 by DeepSeek, achieves gold-level performance at global mathematics competitions.
Chinese AI company DeepSeek has unveiled Math-V2, the first open-source AI model to achieve gold-level performance at the International Mathematical Olympiad.
The system, now available on GitHub and Hugging Face, allows developers to modify and deploy the model under a permissive license freely.
Math-V2 also excelled in the 2024 Chinese Mathematical Olympiad, demonstrating advanced reasoning and problem-solving capabilities. Unlike many AI systems, it features a self-verification process that enables it to check solutions even for problems without known answers.
The launch comes as US AI leaders, such as Google DeepMind and OpenAI, have achieved similar milestones with their proprietary models.
Open access to Math-V2 could democratise advanced mathematical tools, potentially accelerating scientific research and development globally.
Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!
