GPT-5 Sets New Benchmarks Across Every Major AI Leaderboard

OpenAI has released GPT-5, its most capable language model to date. Early benchmarks show significant improvements across coding, scientific reasoning, and multimodal understanding. The model is now available to API customers and is rolling out to ChatGPT subscribers.

Key improvements include near-perfect performance on graduate-level science questions, 40% better code generation accuracy, and native video understanding capabilities.