OpenAI’s GPT-5 Is Here

OpenAI has begun rolling out GPT-5, the latest iteration of its flagship language model, to all ChatGPT users.

The company’s CEO Sam Altman called GPT-5 “a significant step along the path to AGI” during a press briefing on Wednesday. While he stopped short of claiming the model reaches artificial general intelligence, Altman noted the latest release is “clearly a model that is generally intelligent.” He added that GPT-5 still lacks key traits that would make it reach AGI, a notably loose term that is defined in OpenAI’s charter as “a highly autonomous system that outperforms humans at most economically valuable work.” For example, the model still lacks the ability to learn continuously after deployment.

OpenAI claims GPT-5 is smarter, faster, more useful, and more accurate, with a lower hallucination rate than previous models. In characteristically lofty terms, Altman likened the leap from GPT-4 to GPT-5 to the iPhone’s shift from pixelated to a Retina display. “GPT-5 is the first time that it really feels like talking to an expert in any topic, like a PhD level expert,” Altman said.

In ChatGPT, most users will no longer need to choose between models, as the chat interface now automatically routes to the right version depending on the complexity of the query and the user’s subscription tier. As part of Thursday’s launch, OpenAI announced two new model variants: a lightweight gpt-5-mini and an even faster, cheaper, gpt-5-nano (which is only in the API). According to OpenAI, free users will get access to GPT-5 and gpt-5-mini, while the Plus subscription includes the same models with “significantly higher” usage limits. OpenAI says that the $200 a month Pro tier offers unlimited GPT-5 access, along with GPT-5-pro, a more powerful version of the model, and GPT-5-thinking, which allows the model to process a query for longer than usual. Pro users will still have access to pick through legacy models.

Starting next week, Pro users will be able to connect their Gmail, Google Contacts, and Google Calendar to ChatGPT, with other tiers gaining access at an unspecified date. “ChatGPT automatically knows when it’s most relevant to reference them so you don’t need to select them before you chat,” the company said in an email.

Users can also choose a chat color and select from four pre-set personalities—Cynic, Robot, Listener, and Nerd—a feature WIRED’s newsletter Model Behavior reported was in the works last week. According to OpenAI’s blog announcement, it plans to bake these personalities into Advanced Voice Mode.

The company’s API will offer users all three models, along with optional controls for toggling between detailed or direct responses. GPT-5 can retain more information than prior models—it has a 256,000-token context window, a bump up from the 200,000-token context window available in the company’s previous o3 model. That means it can better understand long conversations, documents, or code without losing track of context.

OpenAI claims that GPT-5 beats its previous models on several coding benchmarks, including SWE-Bench (scoring 74.9 percent), SWE-Lancer (GPT-5-thinking scored 55 percent), and Aider Polyglot (scored 88 percent), which test the model’s ability to fix bugs, complete freelance-style coding tasks, and work across multiple programming languages.