OpenAI's new GPT-5 model aims for simplicity and user-friendliness

OpenAI's upcoming releases – GPT-4.5 and GPT-5 – will be focused on delivering a more streamlined and enhanced user experience. That's the promise from CEO Sam Altman, who outlined the company's roadmap in a post on X where he emphasized a shift towards unified AI systems that "just work" for users.

Making AI accessible again: Altman's post makes it clear OpenAI recognizes the complexity of its current product lineup, and plans to simplify its offerings. Acknowledging the complications users face with the existing model selection process, he signaled a return to a more intuitive, unified approach aimed at making AI more accessible and user-friendly.

Chain-of-thought future: GPT-4.5, internally codenamed "Orion," is slated to be OpenAI's last model without chain-of-thought reasoning – where direct answers are generated without intermediate steps, a process that has previously posed challenges in complex reasoning tasks.

Efficiency through versatility: That more advanced process will arrive with GPT-5, which will integrate various OpenAI technologies, including the previously standalone o3 model. Through this integration, OpenAI wants to create a versatile AI system capable of handling a wide range of tasks efficiently.

Adaptable thinking: Altman emphasized that GPT-5 will be adaptable, knowing "when to think for a long time or not," thereby enhancing its utility across diverse applications. While specific release dates have not been announced, Altman indicated that GPT-4.5 and GPT-5 are expected to roll out in the coming weeks and months, respectively.

You get what you pay for: OpenAI plans to offer varied access to GPT-5 across different user tiers. Free ChatGPT users will receive unlimited access to GPT-5 at a standard intelligence setting, subject to abuse thresholds. Plus subscribers (who pay $20/month) will benefit from a higher level of intelligence, while Pro subscribers ($200/month) will access the most advanced features, including voice, canvas, search, and deep reasoning capabilities.