Openai has begun to roll its latest language model, GPT-5 for both free and paid Chatgpt users. The update shows new variants, comprehensive systems integration and performance and improvement in security. So, how do you use chat 5? Well, it is simple, you will automatically show the new Chatgpt-5 welcome screen. This is a staging roll out, so it may take some time for all to be at GPT-5. However, you can always try to rely to check.
GPT-5 is being deployed as a step in the long-term discovery of Artificial General Intelligence of Openai. While the model does not complete the AGI benchmark as defined in its charter of Openai, it is generally described as intelligent. It does not have the ability to learn after deployment.
The company claims that the GPT-5 is more accurate and efficient, with a lower halight rate than earlier versions. Openai CEO Sam Altman said that the jump from GPT-4 to GPT-5 is comparable to the performance upgrade from early iPhones to retinal display.
Two new model variants have been launched: GPT-5-Min, a light version, and GPT-5-Nano, a sharp and cheap model is available only through API. Free users will have access to GPT-5 and GPT-5-Min. Payed levels unlocked GPT-5-PRO with $ 200/Month Pro Tier, GPT-5-Thinking and high use limits with unlimited access.
The Chatgpt interface now selects the most suitable model based on query complexity and user plan. Inheritance models for Pro users are accessible.
On the API Front, GPT -5 is priced at $ 1.25 per million input tokens and $ 10 per million output token. The GPT-5-Mini is priced at $ 0.25 per million input and $ 2 per million output token. GPT-5- Nano is priced at $ 0.05 per million input and $ 0.40 per million output token. The price of Nano variant is now less than Google’s Gemini 2.5 flash.
From next week, Pro users will be able to integrate contacts with Gmail, Google Calendar and Chatgpt. Other levels will be reached later. Chatgpt will automatically refer to these accounts when relevant without manual selection.
Users can now choose from four pricet personalities: Cynic, Robot, Listener and Nerd. These mode is also planned for integration with advanced voice mode. Chat color can also be customized.
The API allows to tolerate between wide and direct reactions. The previous O3 model in GPT-5 has 200,000 to 256,000-token reference window. It enables long interactions, codes and better retention in documents.
GPT-5 to dismiss earlier models in coding works. The benchmark consists of 74.9 percent on SWE-Bench verified, 55 percent on SWE-Lancer using GPT-5-thinking and 88 percent on AIDER polyglot. These cover cover bug fixing, multi-language support and freelance-style tasks.
During a demo, Openai’s team directed the GPT-5 to create an interactive language-education app. The system produced the features and UI requested within a minute. Openai claims that the model can handle the use of complex task chains and tools, including browser and API-based calls.
In the Health Benchmark, GPT-5-Thinking scored 25.5 percent on Healthbench hard than 31.6 percent of O3. Openai states that its new model performs better in health related work, with outputs reviewed by medical professionals.
The company reports that the rate of hallucinations has declined. The GPT-5-Thinking O3 is 65 percent lower rate and 26 percent less than the GPT-4o. Security updates include low potential for model producing misleading, hacked or unprotected outputs. The dual use signals are handled, which the openiAI calls “safe perfection”, provides reactions within the controlled boundaries. More than 5,000 hours of red teaming and external tests were conducted to improve the strength.
The system is designed to thwart safely when faced with an unusual prompt. This involves denying tasks beyond its abilities without generating misleading outputs.
Openai API reports 700 million weekly Chatgpt users, 5 million paying commercial users and 4 million developers.
What’s new in GPT-5: Important change in a glance:
- Available to all users: GPT-5 is being rolled out for both free and paying Chatgpt users.
- New model variants:
- GPT -5 -Mune (Light)
- GPT-5- Nano (API only, fast and cheaper)
- GPT-5-PRO and GPT-5-Thinking (for Pro-Tier users)
- Automatic Model Selection: The chat interface now automatically selects the model based on the complexity of the Query and the user tier. Manual switching is no longer needed.
- Pro-Tier Benefits:
- Unlimited GPT-5 access
- Access to GPT-5-Pro and GPT-5-Thinking
- Heritage models are still available
- Google Integration: Pro users may soon connect Gmail, Google Contact and Google calendar to Chatgpt. Other levels will be found later.
- Adaptation features:
- Users can choose chat color
- Four built -in personality mode: Cynic, robot, listener, NERD
- Later will be integrated with advanced voice mode
- Performance Boost:
- Reference window rose up to 256,000 tokens (above 200,000 in previous O3 model)
- Less hallucination rate compared to GPT-4o and O3
- Better performance in coding benchmarks like Swe-Bench, Swe-Lancer, and Aider Polyglot
- Better in agentic functions and equipment execution (eg, API and browser uses)
- Developer Pricing:
- GPT-5: $ 1.25/1m input token, $ 10/1m output token
- GPT-5-Mini: $ 0.25/1m input, $ 2/1m output
- GPT-5- Nano: $ 0.05/1m input, $ 0.40/1m output
- Health related demonstrations:
- GPT-5- Think O3 O3 in Health Benchmark
- Claimed to be Openai’s best model for medical questions
- Evalted by physicians; Is designed to prostrate on unmatched tasks
- Safety Reforms:
- 26% decrease in hallucinations on GPT-4o
- GPT-5- 65% decrease on O3 for thinking
- Secure perfection for double use
- 5,000+ hours of red teaming and external security tests
- The possibility of misleading or unsafe behavior decreased
