What specific benchmarks highlight Grok 3's superiority over GPT-4o in STEM tasks

1. Mathematics (AIME 2025): Grok 3 achieves a score of 93.3%, significantly outperforming GPT-4o's score of 79% in mathematical reasoning tasks[1][3]. This performance is consistent with Grok 3's ability to solve complex mathematical problems efficiently.

2. Science (GPQA): Grok 3 scores 84.6% on graduate-level expert reasoning, surpassing GPT-4o's score of 78% in scientific problem-solving[1][3]. This indicates Grok 3's advanced capabilities in analyzing scientific data and providing precise answers.

3. Coding (LiveCodeBench): Grok 3 excels with a score of 79.4%, outperforming GPT-4o's score of 72.9% in code generation and programming tasks[1][3]. This showcases Grok 3's ability to generate clean, functional code efficiently.

These benchmarks highlight Grok 3's superior performance in STEM fields, making it a preferred choice for professionals in mathematics, science, and programming. Additionally, Grok 3's Think Mode provides transparency in its reasoning process, which is particularly beneficial for STEM applications[1][3].