New open-source math AI model delivers high performance for just $1,000

Mathletic competition in AI just got hotter. A new open-source AI model for advanced mathematics has emerged, offering high performance with remarkably low training costs. This development represents a significant step in democratizing powerful AI tools, as it provides enterprises and researchers with a freely available, high-performing model that can be modified or deployed for commercial use without restrictions.

The big picture: Researchers have released Light-R1-32B, an open-source 32-billion parameter AI model specifically optimized for solving complex mathematics problems, available on Hugging Face under the permissive Apache 2.0 license.

Performance breakthrough: The new model outperforms similarly sized and even larger open-source alternatives on challenging mathematical benchmarks.

• Light-R1-32B achieved scores of 76.6 on AIME24 and 64.6 on AIME25, surpassing the performance of DeepSeek-R1-Distill-Qwen-32B and DeepSeek-R1-Distill-Llama-70B.
• The American Invitational Mathematics Examination (AIME) contains 15 advanced math problems that human students are given 3 hours to solve.

Cost efficiency: The team completed the model’s training in under six hours using just 12 Nvidia H800 GPUs, with a total estimated cost of only $1,000.

• This represents one of the most economical approaches for developing high-performing specialized AI models.
• The rapid training time demonstrates the potential for more accessible AI development in specialized domains.

Technical foundation: Light-R1-32B builds upon Alibaba‘s open-source Qwen 2.5-32B-Instruct model.

• The researchers have also released their training datasets, scripts, and evaluation tools alongside the model.
• Despite its mathematics specialization, the model maintains strong generalization abilities on scientific reasoning tasks.

Why this matters: The permissive Apache 2.0 license allows organizations to use, modify, and commercialize the model without restrictions, potentially accelerating innovation in AI-powered mathematics and scientific applications.