Gemini 4 Carbon takes the MathArena crown but costs 15 times more than GPT-6.1 Sol

Google's Gemini 4 Carbon model has surpassed OpenAI's GPT-6.1 Sol to top the MathArena math benchmark, achieving 85.4% accuracy but at a cost about 15 times higher per query. Anthropic's Claude Fable 5.2 ranked second, with the highest cost among the top models. While OpenAI's models have demonstrated solving complex equations, they remain unavailable to the public.