OpenAI on Thursday launched GPT-5, its most advanced artificial intelligence model to date, making it available to all ChatGPT users, including those on the free tier. The company said the model is faster, smarter and "a lot more useful," with improvements in reasoning, coding and specialized tasks such as health care.
During a conference call with journalists, OpenAI CEO Sam Altman described GPT-5 as a "major upgrade" and "a significant step along the path of AGI". He added that after trying GPT-5, "going back to GPT-4 was miserable." The model will be offered in three versions-standard, mini and nano-tailored for different cost and latency requirements.
The release comes as ChatGPT approaches 700 million weekly active users, up from 500 million in April. OpenAI is reportedly in talks with investors for a stock sale valuing the company at about $500 billion. Microsoft, a key partner, said GPT-5 will be integrated into Microsoft 365 Copilot, consumer Copilot, and Azure AI Foundry starting Thursday.
OpenAI said GPT-5's hallucination rate-instances in which the model fabricates information-is lower than prior versions. The company conducted 5,000 hours of safety testing and introduced "safe completions," which allow the model to give constrained, high-level responses to potentially risky prompts rather than refusing outright.
"GPT-5 has been trained to recognize when a task can't be finished, avoid speculation and can explain limitations more clearly," said Michelle Pokrass, post-training lead at OpenAI. Altman likened the new model's capabilities to "having a team of Ph.D.-level experts on hand at any time."
The model's enhanced coding capabilities were showcased in a live demonstration of "vibe coding," in which GPT-5 generated two fully functional French-learning web apps from identical prompts in seconds. While the apps required minor adjustments, OpenAI said users could easily customize elements such as backgrounds and navigation tabs.
Pro and Team subscribers will have unlimited access to GPT-5, while Plus users receive higher usage limits. Free users can use the full reasoning model until hitting their cap, at which point they are switched to GPT-5 mini. ChatGPT Edu and Enterprise customers will gain access within a week.
Enterprise testing is already underway. Aaron Levie, CEO of cloud content company Box, said GPT-5 passed advanced internal tests where earlier models had failed. "The model is able to retain way more of the information that it's looking at, and then use a much higher level of reasoning and logic capabilities to be able to make decisions," he said.
Earlier this week, OpenAI also released two open-weight language models-the first since GPT-2 in 2019-designed for developers and researchers seeking lower-cost, customizable options. But with GPT-5, Altman emphasized the company's focus on mainstream adoption, saying interacting with the model "feels natural and more human."
