OpenAI chief executive officer Sam Altman has given the first public glimpse of the company’s next flagship model, posting a screenshot of what he described as “GPT-5” responding to a question about artificial-intelligence-themed television. The image, confirmed by Business Insider, signals that the long-anticipated successor to GPT-4 is close to release and is expected to offer a larger context window, stronger reasoning abilities and multimodal features. Developer chatter, leaked code and early tester reports suggest OpenAI may unveil a family of models—including smaller GPT-5-mini and GPT-5-nano versions—and its first open-weights systems since GPT-2, at roughly 120-billion and 20-billion parameters. The company is also preparing a lower-priced “ChatGPT Go” subscription intended to sit below the current US$20-per-month Plus plan, according to references found in the ChatGPT web app. In an interview published by MIT Technology Review, OpenAI research heads Mark Chen and Jakub Pachocki said recent internal reasoning models have already reached medal-level scores in international math contests and placed second in a 10-hour coding marathon, achievements they view as building blocks for GPT-5. External testers report that the new model can refactor large codebases and tackle complex academic problems, underscoring OpenAI’s push to extend autonomous problem-solving while it races against rivals such as Google DeepMind, Meta, xAI and Anthropic.