GPT-5 is a multimodal large language model (LLM) developed by OpenAI, serving as the flagship successor to the GPT-4 series. Released in August 2025, it was designed as a unified system that integrates high-speed processing with deep reasoning capabilities, effectively merging features of the previous GPT and o-series models. The model is available in several variants, including GPT-5, GPT-5 Mini, and GPT-5 Nano, alongside specialized "thinking" versions optimized for complex logical tasks.
Technically, GPT-5 supports a significantly expanded context window of up to 400,000 tokens via its API, enabling the processing of extensive documents and long-horizon tasks. The model architecture emphasizes improved factuality, advanced coding proficiency, and native multimodal support across text, image, audio, and video. It introduced new developer controls, such as adjustable reasoning effort and verbosity, allowing for more granular optimization of compute resources and response detail.
OpenAI also utilized the GPT-5 launch to release GPT-OSS, a set of open-weight models with reasoning capabilities, marking a shift toward transparency for specific use cases. However, the flagship GPT-5 remains a proprietary model integrated into the ChatGPT ecosystem and the OpenAI API, featuring enhanced safeguards and a system card that outlines its capabilities in scientific reasoning and agentic workflows.