The latest model demonstrates human-level performance across scientific reasoning, coding, and multimodal understanding tasks.
OpenAI has officially released GPT-5, its most powerful language model to date. The model achieves human-level performance on a wide range of benchmarks including MMLU, HumanEval, and MATH.
GPT-5 introduces a new architecture that combines chain-of-thought reasoning with multimodal understanding. The model can process text, images, audio, and video simultaneously.
In internal evaluations, GPT-5 scored 96.8% on MMLU, surpassing all previously published models. On coding benchmarks, it solves 87% of competitive programming problems.
GPT-5 is available via the OpenAI API starting today, with ChatGPT Plus users getting access within the next 48 hours.
Original source
OpenAI Blog
Share this article
No community notes yet
Be the first to add a technical insight.