OpenAI’s (OPAI.PVT) GTP-5 is officially available. The company’s most advanced AI model yet, and available now at ChatGPT.com, OpenAI says GPT-5 is a step up from its prior GPT models across a number of areas, including writing, coding, health knowledge, and more.
According to CEO Sam Altman, GPT-5 is like speaking to a “legitimate PhD-level expert.” It’s also, the company says, less prone to hallucinations than previous releases.
The biggest change you’ll notice when using GPT-5 is that it is the only model you can use. You’ll no longer be able to use GPT-4o or any other GPT version. But OpenAI says you shouldn’t be too concerned about that.
That’s because GPT-5 can automatically switch between providing quick, efficient answers to queries and taking time to reason longer over more complex questions, depending on how you ask them. It’s a helpful feature, to be sure, since you don’t have to take the time to decide whether you’re using the right model for your particular ask.
If you’re a Plus or Team user, you also have the option of selecting GPT-5 Thinking as your base model. But free users are out of luck. When you run through your usage limit, ChatGPT will automatically switch to smaller, less resource-intensive GPT-5 models.
According to OpenAI, GPT-5 offers enhanced writing skills that include more literary depth and rhythm. It can also help you translate information from things like complex medical reports into plain English, making them far easier to understand. Of course, the model isn’t bulletproof, so you’ll want to double-check any medical information and run it by your doctor before making any decisions.
GPT-5 is also more adept at coding than earlier GPT models. It can put together things like basic games and apps with just a few prompts and details. And unlike previous versions of GPT, GPT-5 gives your apps a bit of stylistic flair, rather than making them feel dull and flat.
That should prove helpful for the types of folks who are interested in “vibe coding,” or writing software using natural language prompts rather than actual code.
GPT-5 is also more personal thanks to its ability to connect to Google (GOOG, GOOGL) Calendar, Gmail, and Contacts. In one demo, OpenAI showed how GPT-5 can help pull together a schedule based on your calendar and travel information from your email. It’s the kind of helpfulness that is certain to attract some users, while turning off others who prefer to keep their more sensitive data private and away from third-party apps.
Importantly, OpenAI says GPT-5 is less sycophantic than its predecessors. The company originally released an update to GPT-4o in April but quickly pulled it back after the chatbot began being overly flattering to users. It also, the company said, was validating users’ doubts, fueling anger, and urging impulsive actions.
Reports indicate that some users can experience emotional spirals when interacting with overly sycophantic bots. That should change with GPT-5.
“In targeted sycophancy evaluations using prompts specifically designed to elicit sycophantic responses, GPT‑5 meaningfully reduced sycophantic replies (from 14.5% to less than 6%),” the company said in its release announcing the new model.
“At times, reducing sycophancy can come with reductions in user satisfaction, but the improvements we made cut sycophancy by more than half while also delivering other measurable gains, so users continue to have high-quality, constructive conversations — in line with our goal to help people use ChatGPT well.”
While GPT-5 is an important release, it won’t put OpenAI in the AI driver’s seat for long. Its competitors, including Anthropic (ANTH.PVT), Google, Meta (META), and Perplexity (PEAI.PVT), working on their own next-generation AI models. And in the ongoing tit-for-tat battle for AI supremacy, no company stays on top for long.
