GPT-5 vs Gemini: Inside 2025’s High-Stakes Frontier AI Showdown
Gemini Ultra became the first AI to surpass human expert performance on the MMLU benchmark with a 90% score. Meta released LLaMA 4 in April 2025, reportedly featuring over 400 billion parameters. OpenAI’s O1 model scored 83% on the IMO math test, far ahead of GPT-4o’s 13%. Fifteen organizations signed the Frontier AI Safety Commitments at the 2024 Seoul Summit.