OpenAI has officially unveiled GPT-5, revealing more than just a new AI model. The presentation featured a redesigned model offering, new customization options, deeper integrations with Google services, security improvements, and expanded capabilities for developers. With this, the company is marking another milestone in the development of ChatGPT, creating a cleaner, more versatile platform.
The event, which was broadcast live for an unusually long time, highlighted OpenAI's increasing focus on user-friendliness and application diversity. Instead of focusing solely on performance improvements, the company presented a significantly streamlined model portfolio, standardized features for both free and Plus users, expanded developer tools, and a stronger focus on responsible use. It also became clear that GPT-5 doesn't make giant leaps in all areas, but has been specifically improved where it matters most in everyday life and in practice.
A model for everyone
OpenAI has reduced its previously complex model offering to three main variants: GPT-5, GPT-5 Thinking, and GPT-5 Pro. The Pro version is reserved for the $200 monthly plan. ChatGPT will now automatically decide how much computing time will go into a response. Those who want to force a longer reasoning process can manually select GPT-5 Thinking. Free users will have access to the same model as Plus subscribers for the first time, but with stricter usage limits. Once these limits are reached, ChatGPT automatically switches to a Mini version. Plus users benefit from more generous limits but must also adhere to this rule. All older models will be discontinued. Existing conversations will automatically be converted to the appropriate GPT-5 equivalent: Chats with 4o, 4.1, 4.5, 4.1-mini, o4-mini, or o4-mini-high will open in GPT-5, chats with o3 will open in GPT-5 Thinking, and chats with o3-Pro will open in GPT-5 Pro. The voice mode will remain in GPT-4o for the time being.
Voice mode and personalization
Voice mode can now also be used with custom GPTs. Paying users can precisely specify how long, how fast, and in what tone answers are given. There are also four new predefined personalities in addition to the standard version, which can also be customized. Cynic delivers sarcastic, dry answers, often humorous, but direct. Robot gives precise, unemotional answers without unnecessary embellishments. Listener sounds warm and relaxed, conveying content clearly and with light humor. Nerd appears playful and inquisitive, explaining concepts clearly. The user interface can also be personalized with accent colors.
Google Connectivity
Starting next week, Pro users will be able to connect their Google Calendar, Google Contacts, and Gmail accounts to ChatGPT to receive personalized support. In the live demo, an OpenAI engineer used the integration to organize appointments, with ChatGPT even autonomously suggesting a jog between meetings. Expansion to other user levels is planned, although there is no timeline yet.
Safety and health
With Safe Completion, OpenAI is addressing security concerns. Instead of directly rejecting risky inputs, GPT-5 now explains why a request cannot be continued. The company aims to become more transparent and avoid potential misunderstandings. In the healthcare sector, OpenAI emphasizes a significantly lower hallucination rate compared to o3 and GPT-4o. During the event, a conversation between CEO Sam Altman and a cancer patient and her husband was shown. They described how GPT-5 helped them better understand the disease and treatment. OpenAI makes it clear that medical answers from AI should not be blindly trusted, but sees the responsible explanation of information as an important step forward.
Developer features
GPT-5 offers new possibilities for developers. In addition to the main model, the GPT-5 Mini and GPT-5 Nano variants are available via the API. The context window has been expanded from 200,000 tokens (o3) to 256,000 tokens. Prices are $1.25 per million input tokens and $10 per million output tokens for GPT-5, $0.25 and $2 for Mini, and $0.05 and $0.40 for Nano. The verbosity of responses can now be customized, allowing for more precise control over token usage. In a live presentation, the CEO of Cursor demonstrated how GPT-5 can be integrated into an AI-powered IDE to solve programming tasks faster and more efficiently. OpenAI also demonstrated the rapid creation of a language learning game and an interactive demo on aircraft aerodynamics.
GPT-5 combines stability, customization and smart integration
GPT-5 represents more than just performance improvements. OpenAI focuses on clear structures, increased customization, integration into existing workflows, and improved security. While the performance leaps are smaller compared to previous versions, GPT-5 offers noticeable improvements in reliability, adaptability, and developer-friendliness. With the combination of new features and a simplified model range, OpenAI strengthens its position as one of the leading providers of modern AI technology. The best products for you: Our Amazon Storefront offers a wide selection of accessories, including those for HomeKit. (Image: Shutterstock / Azulblue)
- OpenAI releases GPT-5: Fewer errors, more performance
- OpenAI brings free language models for local use
- New ChatAI app: Bragi and OpenAI combine their strengths
- ChatGPT Agent: OpenAI now lets AI solve tasks itself
- ChatGPT for Mac: Recording mode now available for Plus users
