OpenAI releases GPT-5, the "PhD-level expert" chatbot
Sam Altman and his artificial intelligence company OpenAI have introduced their latest flagship model, GPT-5, the most intelligent chatbot in existence. It can literally write entire computer programs in one prompt.
Image by Cybernews
Sam Altman and his artificial intelligence company OpenAI have introduced their latest flagship model, GPT-5, the most intelligent chatbot in existence. It can literally write entire computer programs in one prompt.
- OpenAI introduces GPT-5, a more robust large language model with expert-level intelligence and reasoning skills.
- The more adaptable, honest, and less hallucinogenic GPT-5 is available for free to all users starting today.
- Highlights include advanced coding abilities and improved factuality for myriad subjects, acheiving the highest level of AI benchmarking.
CEO Sam Altman kicked off the one-hour-plus livestream event on the OpenAI YouTube channel and its X profile on Thursday at 10:00 a.m. Pacific Time/1:00 p.m. Eastern Time.
Calling the chatbot “a significant step along our path to AGI” Altman boasted that the “useful, smart, fast, and intuitive” GPT-5 is a major upgrade over the company’s previous model.
The San Francisco-based AI start-up and over a dozen top team members demoed GPT-5’s most exciting features, showing off its “state-of-the-art performance across coding, math, writing, health, and visual perception.”
GPT-5 is here.
undefined OpenAI (@OpenAI) August 7, 2025
Rolling out to everyone starting today.https://t.co/rOcZ8J2btI pic.twitter.com/dk6zLTe04s
OpenAI is rolling out a basic version of GPT-5 to all its 700 million users for free on Thursday. However, free users will be given a limit on the number of prompts allowed.
When a user hits the so far undisclosed limit, they'll be automatically transitioned to GPT-5 mini, "a smaller but still highly capable model," OpenAI said.
Plus subscribers are being granted higher usage limits, and Pro subscribers are getting access to a GPT-5 Pro model – said to provide users with “even more comprehensive and accurate answers.” The Enterprise and Edu GPT-5 model versions will be available next week.
Smarter than the smartest person
“Someday soon, something smarter than the smartest person you know will be running on a device in your pocket, helping you with whatever you want. This is a very remarkable thing,” Altman teased earlier this week.
As of its release, GPT-5 now holds the highest Chatbot Arena score to date, outshining Gemini 2.5, Grok 4, and all other existing large language model competitors.
“It's like talking to an expert, a legitimate PhD-level expert in anything,” Altman said.
“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,” the company said.
Furthermore, when measuring reasoning for economically valuable work, OpenAI said GPT‑5 is comparable to or better than experts in about 50% of cases across tasks in “over 40 occupations, including law, logistics, sales, and engineering.”
Solving the reasoning paradigm
The “incredible superpower on demand” is said to be light on hallucinations and heavy on reliability when it comes to reasoning complex, open-ended questions.
Before GPT-5, "our users have had to pick between the fast responses of standard GPT or the slow, more thoughtful responses from our reasoning models,” said OpenAI’s Chief Research Officer Mark Chen.
“GPT-5 eliminates this choice, aiming to think just the perfect amount to give you the perfect answer,” Chen said.
In one demo shown, GPT-5 was instructed to answer the user's query with either one-word, concise, or more expanded responses, and for those using speech prompts, even allowing the user to choose more personalized GPT options to receive their answers.
Users will also get to choose their chatbot’s communication style from four options in the settings menu – the Cynic, the Robot, the Listener, and the Nerd, of course. The personality descriptions? Concise and professional, thoughtful and supportive, or a bit sarcastic.
GPT-5 has 4 new chat personalities: Cynic, Robot, Listener, Nerd.
undefined OpenAI (@OpenAI) August 7, 2025
Find them in Customize ChatGPT in settings.
Research preview. Text-only. Opt-in. Change anytime.
Written by Robot in GPT-5. pic.twitter.com/XxyWYNYxHt
Eventually coming to voice, the feature is only available in text at the moment.
Computer programming on demand
The OpenAI team said since the launch of Chat-GPT in November 2022 they have gained “a much better understanding of how people actually want to work with chat,” allowing developers to “optimize” GPT-5 for use cases, such as coding.
“We've come a long way from the days when, you know, only 5 to 10 lines of code were working, and now it's amazing that you can produce these kinds of apps on demand,” said Chen.
Calling GPT-5 “the best coding model on the market today,” Altman predicts the idea of “software on demand” will be one of the defining characteristics of the GPT-5 era.
GPT-5 proves it can quickly create websites, apps, and games, with early testers reporting better design choices and understanding of visual aesthetics such as spacing, typography, and white space.
In a demonstration, OpenAI researcher Yann Dubois showed off GPT-5 capabilities, writing code to create an app based on only one prompt description.
Dubois, who is a native French speaker, asked GPT-5 to build a web app for his partner to learn how to speak French.
Here is pretty much the exact prompt:
“Create a beautiful and highly interactive web app for my partner, an English speaker at Tulane (University) French. Use a highly engaging theme. Include a variety of activities, like flashcards and quizzes that she can interact with. To make it even more fun for her, embed an educational game based on the old snake game, but replace the snake with a mouse and the apples with cheese. Every time the mouse eats a piece of cheese, voice over a new French word so my partner can practice her pronunciation.”One prompt and GPT-5 can build an app on demand.
Barely two minutes later, GPT-5 “already wrote 240 lines of code, which is honestly much more than what I would have written in that time,” Dubois says, and then proceeds to run the code, showing off three different working versions of the newly created app.
GPT-5 "opens up a whole new world of vibe coding… even those who do not know how to write code can bring their ideas to life,” Dubois said. The new model is also equipped with upgraded writing capabilities, which make it more collaborative and reliable and possess more “literary depth and rhythm.”
Finally, GPT-5 is considered the best model for health-related questions, surpassing all other models previously reached benchmarks.
Acting like a health advocate or partner, the chatbot can provide the user with more in-depth medical knowledge, helping individuals understand diagnoses and make more informed decisions.
Earlier this week, Altman announced on X that OpenAI will now provide “ChatGPT access to the entire federal workforce” at the bargain basement price of only $1 per year for each government agency.
FAQ
Is GPT-5 better than GPT-4?
According to OpenAI, yes, a lot better. The company claims GPT-5 to be the best chatbot yet, as it is capable of performing better when needed to perform tasks like reasoning, being factually accurate, and generally increasing its task performance. In comparison to GPT-4o, GPT-5 is roughly 45% less likely to make factual errors and can deliver PhD-level responses on subjects such as engineering, law, and health. OpenAI claims to have also developed a better response time to prompts that require deep reasoning, solving the earlier trade-off users faced between speed and thoughtfulness.
How much will ChatGPT-5 cost?
Access to GPT-5 depends on the user’s plan. At first, it is indeed free, but there is a limited number of prompts that a user can use. The exact number is not yet disclosed, but once a user reaches it, they could try out the GPT-5 mini, a lighter GPT-5 version. According to Wired, which cites a “developer launch blog,” the cost of GPT-5 will depend on the user's version – GPT-5, GPT-5 mini, or GPT-5 nano – and on the number of input and output tokens processed through the API. “GPT-5 mini is priced at $0.25/1M input tokens and $2/1M output tokens, and GPT-5 nano is priced at $0.05/1M input tokens and $0.40/1M output tokens,” it adds. Developers do not use GPT-5 directly. Instead, an API (Application Programming Interface) allows them to integrate it into their own services, such as a support chatbot in an online shop. When a client asks a question to the bot, it becomes a prompt that the GPT-5 resolves and ”gives back” in the form of an answer. Therefore, the cost is based on the number of tokens processed (they are separately priced for what’s sent) and the output that GPT-5 generated. For users who are not developers, ChatGPT Plus will be $20 per month.
How is GPT-5 different from earlier versions?
GPT-5 introduces several upgrades: Expert-level knowledge across over 40 professional fields, such as the aforementioned law or medicine. Advanced coding capabilities – it can now build full apps, games, and websites from a single prompt. Users can choose their communication styles from options such as “The Listener,” “The Nerd,” and “The Cynic.” Improved writing that offers deeper rhythm and collaboration for creative work. Reduced hallucinations and more accurate reasoning, even on open-ended and complex questions.