The recent advancements in mathematical reasoning and problem-solving using large language models (LLMs) and multi-modal large language models (MLLMs) have shown significant progress in handling complex mathematical tasks. The field is moving towards more sophisticated models that integrate visual and textual data for enhanced understanding and reasoning. Key developments include the introduction of models capable of solving geometry problems, generating LaTeX equations from speech, and automating the discovery of mathematical constants and their relations. Notably, there is a growing emphasis on creating models that not only solve problems but also propose and prove theorems, akin to a 'geometry coach'. Additionally, the integration of coding instruction into LLMs for mathematical reasoning has shown promising results, with models demonstrating improved performance through diverse coding styles. The democratization of advanced mathematical systems, such as those capable of handling Olympiad-level problems, is also a notable trend, making high-level mathematical reasoning more accessible. Furthermore, the field is witnessing a shift towards more comprehensive and robust benchmarks for evaluating mathematical reasoning capabilities, with a focus on creating datasets that better represent mathematical research practices and proof-discovery processes. These developments collectively indicate a move towards more integrated, accessible, and advanced mathematical reasoning tools.