OpenAI has launched GPT-5, its newest and most advanced large-scale artificial intelligence model, making it immediately available to all ChatGPT users, including those on the free tier. The company asserts that GPT-5 is smarter, faster, and significantly more useful, particularly in areas such as writing, coding, and healthcare.
OpenAI CEO Sam Altman highlighted the substantial leap in capability, stating in a briefing with reporters, “I tried going back to GPT-4, and it was quite miserable.” This sentiment underscores the perceived difference in performance between the new model and its predecessor. Since the debut of its AI chatbot ChatGPT in 2022, OpenAI has experienced rapid mainstream adoption. The company anticipates reaching 700 million weekly active users on ChatGPT this week. Furthermore, OpenAI is reportedly engaged in discussions with investors regarding a potential stock sale that could value the company at approximately $500 billion.
A key enhancement in GPT-5 is its reduced hallucination rate, meaning the model fabricates answers less frequently. OpenAI also emphasized the extensive safety evaluations undertaken during GPT-5’s development, including 5,000 hours of rigorous testing. Instead of outright refusing to answer potentially risky user questions, GPT-5 employs “safe completions.” This approach allows the model to provide high-level responses within established safety constraints, ensuring the information cannot be used to cause harm. Michelle Pokrass, a post-training lead at OpenAI, explained, “GPT-5 has been trained to recognize when a task can’t be finished, avoid speculation and can explain limitations more clearly, which reduces unsupported claims compared to prior models.”
During the briefing, OpenAI showcased GPT-5’s “vibe coding” capabilities, a term referring to the generation of software using AI based on simple written prompts. In a demonstration, the company tasked GPT-5 with creating a web application designed to help an English speaker learn French. The prompt specified an engaging theme, inclusion of activities like flashcards and quizzes, and a feature for tracking daily progress. OpenAI submitted the same prompt to two separate GPT-5 windows, and within seconds, the model generated two distinct applications. While acknowledging “some rough edges,” an OpenAI lead noted that users retain the flexibility to make further adjustments to the AI-generated software, such as altering the background or adding tabs, as needed.
GPT-5 began rolling out to OpenAI’s Free, Plus, Pro, and Team users on Thursday. This release marks a significant milestone for Free users, as it grants them access to a reasoning model for the first time. Reasoning models are designed to “think” or execute an internal chain of thought before formulating a response. Free users who reach their usage cap will still have access to GPT-5 mini. OpenAI’s Plus users benefit from higher usage limits, while Pro users enjoy unlimited access to both GPT-5 and GPT-5 Pro. ChatGPT Edu and ChatGPT Enterprise users are slated to gain access to GPT-5 approximately one week from Thursday.
The integration of GPT-5 extends beyond OpenAI’s direct offerings. Microsoft CEO Satya Nadella, in a Thursday X post, reflected on the rapid progress since Sam Altman’s February 2023 GPT-4 demonstration in Bing. Nadella stated, “It’s hard to believe it’s only been two and a half years since @sama joined us in Redmond to show the world GPT-4 for the first time in Bing, and it’s incredible to see how far we’ve come since that moment.” According to a Microsoft blog post, the new model is being integrated into various Microsoft products starting Thursday, including Microsoft 365 Copilot, the Copilot for consumers, and Azure AI Foundry, which enables developers to incorporate AI models into third-party applications.
Box, an enterprise file management company, has been testing GPT-5 across diverse datasets in recent weeks. Aaron Levie, CEO of Box, praised GPT-5 as a “complete breakthrough.” Levie observed that previous AI models often struggled with Box’s advanced tests involving complex math or logic within lengthy documents. He told CNBC, “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.”
For developers, OpenAI is releasing three different versions of the model via its application programming interface (API): gpt-5, gpt-5-mini, and gpt-5-nano. These versions are tailored to meet varying cost and latency requirements. Earlier in the week, OpenAI also released two open-weight language models, marking the first time since GPT-2 in 2019. These open-weight models are intended to provide lower-cost options that developers, researchers, and companies can easily run and customize.
With GPT-5, OpenAI aims for a more natural and “human” interaction experience for a broader consumer audience. Altman likened GPT-5 to having a team of Ph.D.-level experts readily available at any time. He concluded, “People are limited by ideas, but not really the ability to execute, in many new ways.”
