MathBlackBox is a deep learning model designed to explore black-box methods for solving mathematical problems. It utilizes VLLM or other OpenAI-compatible approaches, conducts inference through the Huggingface toolkit and OpenAI, supports operation within Slurm environments, and can process various datasets. This project is currently in its early stages and requires thorough testing before deployment in real-world applications.