OpenAI to Release GPT-5.6 in Phases Per US Government Request

The US government has asked OpenAI to release its next-generation model GPT-5.6 in phases due to security concerns. CEO Sam Altman confirmed in a company Q&A that the model will launch as a limited preview, accessible only to select enterprise customers, with each access approved by the US government on a case-by-case basis. This is more favorable than the conditions for rival Anthropic, which was told to pause its models. Some Pro subscribers can already access GPT-5.6 Pro, with significantly improved output quality. The model's context window has expanded from 1 million to 1.5 million tokens, enhancing long coding tasks and response speed. In agent coding, GPT-5.6 outperforms Anthropic's Mythos series, becoming a new benchmark. OpenAI plans to cut prices further while improving performance, with current token costs about half of Anthropic's. OpenAI also plans to wait until next year (2027) for its IPO.