OpenAI, the company behind the popular generative AI chatbot ChatGPT, has released GPT-5, with three variants — GPT-5 Pro, GPT-5 mini, and GPT-5 nano. The San Francisco-based company claims this is the “best AI system yet”, and it comes with reduced chances of hallucinations, more powerful coding capabilities, and a new approach to handling sensitive requests or “safe completions”.
Pieces of the puzzle that make an artificial intelligence model perfect are beginning to come together. It allows “state-of-the-art performance across coding, math, writing, health, and visual perception”, offering a “significant step forward,” but the company CEO Sam Altman said, “we’re still missing something quite important” as GPT-5 can’t continuously learn on its own.
He said GPT-5 offers a considerable leap over previous models, “something that I just don’t wanna ever have to go back from”.
The company said in a statement that GPT‑5 is “significantly less likely to hallucinate than our previous models”. With web search enabled on anonymised prompts representative of ChatGPT production traffic, “GPT‑5’s responses are ~45% less likely to contain a factual error than GPT‑4o, and when thinking, GPT‑5’s responses are ~80% less likely to contain a factual error than OpenAI o3”.
Altman said: “GPT-4 felt like you’re kind of talking to a college student. GPT-5 is the first time that it really feels like talking to an expert in any topic, like a PhD-level expert.”
GPT-5 in an integrated model, “meaning no more model switcher and it decides when it needs to think harder or not”. It is available to everyone, including the free tier, with reasoning.
OpenAI claims that it has trained GPT-5 to “honestly” communicate “its actions and capabilities to the user — especially for tasks which are impossible, underspecified, or missing key tools”.
The new model is quick with responses and better at answering queries. Nick Turley, who runs OpenAI’s ChatGPT team, said: “When you’re talking to this thing, it feels just a little bit more natural.” Further, ChatGPT can decide when to spend more time thinking over a user’s query, offering users the best response.
OpenAI research Tina Kim said during the event that the company is “deprecating all of our old models” or shutting down an obsolete model.
Demonstrations have shown that it is zippy while creating custom applications with no coding involved. Yann Dubois, a post-training lead at OpenAI, asked the model to create an app for learning French. It took only a few minutes to come up with a web application complete with sound and working game functions.
There are also safety features, which come into play when a dangerous prompt is put out. Earlier, OpenAI models would refuse to answer such prompts, but GPT-5 tries to offer the best safe answer. Harmful questions may come in the form of an innocent chemistry question, but may have a harmful consequence, like building weapons.
The company is launching a research preview of four new preset personalities for all ChatGPT users. These personalities, available initially for text chat and coming later to Voice, let users set how ChatGPT interacts—whether concise and professional, thoughtful and supportive, or a bit sarcastic — without writing custom prompts. The four initial options, Cynic, Robot, Listener, and Nerd, are opt-in.
Starting next week, Pro users can connect their Gmail, Google Contacts, and Google Calendar to ChatGPT. Once done, you can ask the chatbot about your schedule, and it will offer suggestions.
GPT-5 is arriving at a time when competition has intensified from Alphabet’s Google, Anthropic, and xAI, as well as the Chinese startup DeepSeek. Further, Meta Platforms is spending heavily to hire top talent for its new Superintelligence Lab, including poaching more than a dozen employees from OpenAI.
The company’s CEO said: “GPT-5 is the smartest model we’ve ever done, but the main thing we pushed for is real-world utility and mass accessibility/affordability. We can release much, much smarter models, and we will, but this is something a billion-plus people will benefit from.” As a footnote, he added: “Most of the world has only used models like GPT-4o.”
Altman said the model is a “significant step” towards achieving AGI or artificial general intelligence, which refers to models that are as capable as a human. But he also said the model is not there yet because it can’t learn continuously. Meta CEO Mark Zuckerberg has been recruiting AI scientists intending to achieve “superintelligence”.
GPT‑5 is starting to roll out immediately to all Plus, Pro, Team, and Free users, with access for Enterprise and Edu coming in one week. Pro, Plus, and Team users can also start coding with GPT‑5 in the Codex CLI by signing in with ChatGPT.
OpenAI is set to reach 700 million weekly active users for ChatGPT this week, up from 500 million in March, representing a more than fourfold year-over-year surge in growth.
Microsoft and GPT-5
Microsoft, which is an investor in OpenAI, is bringing GPT-5 to Copilot, Microsoft 365 Copilot, Azure AI Foundry, GitHub Copilot, and more.
Developers across Microsoft platforms will have access to OpenAI’s GPT-5 as well. As an example, Microsoft cited: “Developers using GitHub Copilot and Visual Studio Code can write, test and deploy code using OpenAI’s best model yet for coding and agentic tasks, which OpenAI says can complete longer and more complex coding tasks and excels at executing long-running agentic tasks end-to-end.” Developers will also have access to the latest GPT-5 models in Azure AI Foundry, with trusted enterprise-grade security, compliance and privacy protections available in Azure.
Microsoft 365 Copilot users also have access to GPT-5. “With GPT-5, Microsoft 365 Copilot is better at reasoning through complex questions, staying on track in longer conversations and understanding the user’s context,” said Microsoft in a blog post.
Microsoft Copilot is the AI companion for everyday tasks. Users can experience the power of GPT-5 in Copilot for free.