GPT-5 combines fast answers and deeper reasoning in ChatGPT
OpenAI launches GPT-5 as a routed system that chooses between quick responses and longer reasoning, with stronger coding and lower factual-error rates.
OpenAI launched GPT-5 today in ChatGPT and its programming interface. ChatGPT uses a routing system that can answer quickly or invoke a deeper reasoning model according to the request, its complexity and a user's explicit instruction.
The company reports gains in coding, writing, visual understanding and health questions, along with fewer factual errors than earlier systems in its evaluations. Developers receive three model sizes and controls over reasoning effort and answer length, allowing applications to balance capability, speed and cost.
A unified product can spare users from choosing among a confusing menu of models, although the router itself becomes part of the experience and may not always allocate enough thought. Better benchmark results also leave the familiar obligation to verify consequential answers. The most useful change may be practical rather than theatrical: stronger instruction following and tool use could let the model complete longer pieces of real work with fewer corrections.
Source: OpenAI ↗