After months of anticipation, OpenAI has officially released GPT-5, its most advanced large language model to date. The new model, which is now available for both free and paid users, promises across-the-board improvements in writing, coding, reasoning, and safety.
Read: Bandwidth Blog & Smile 90.4FM Tech Tuesday: iPad Pro and VR?
In a press briefing, OpenAI CEO Sam Altman described the experience of using GPT-5 as “talking to an expert in any topic,” comparing the leap in capability to the iPhone’s transition to a Retina Display.
GPT-5 delivers on Altman’s promise of a unified user experience. For the first time, OpenAI’s default model is a reasoning model, designed to tackle complex problems by breaking them down into smaller parts. This simplifies the user experience, as the previous “Think Longer” option for reasoning models is no longer necessary.
On the consumer side, OpenAI has streamlined its offerings. The main model is GPT-5, but Plus subscribers can use it “significantly” more than free users. A new GPT-5 mini model will be used when free and Plus users hit their query limits, while a top-tier GPT-5 Pro is available for a new $200 per month subscription.
GPT-5 is not only more powerful than its predecessor, it’s also faster. Altman noted that its speed is so impressive that it can feel like it’s not “thinking enough,” even as it provides excellent answers.
A major focus of GPT-5 is reducing hallucinations, a common weakness in previous reasoning models. OpenAI claims the new model is 80% less likely to include a factual error than its previous state-of-the-art model, o3.
The company also says GPT-5 is its safest model yet, thanks to a new feature called Safe Completions. This feature allows the model to provide helpful, high-level information in potentially tricky situations without compromising user safety.
See also
GPT-5 is launching with a host of new features for ChatGPT:
- Google Integration: You can now easily connect ChatGPT to your Google accounts (Gmail, Calendar, Contacts), allowing the chatbot to automatically access relevant information. This feature is rolling out to Pro users first.
- Customization: Users can now select different chat colours, with exclusive options for paid subscribers.
- Pre-set Personalities: A new research preview offers four pre-set personalities—Cynic, Robot, Listener, and Nerd—that you can use to tweak how the chatbot interacts with you.
- Advanced Voice: An updated version of the Advanced Voice feature is more adept at understanding instructions and adapting its speaking style. The less capable Standard Voice Mode has been retired.
When asked about the path to artificial general intelligence (AGI), Altman was cautious. “We’re still missing something quite important,” he said, noting the model’s inability to learn continuously on its own. However, he described GPT-5 as a “significant step forward” and a “significant fraction of the way to something very AGI-like.”
