Key takeaways of zdnet
- Openai has launched its long-awaited GPT-5 model.
- The model claims that the fastest, smartest and most capable of OpenaiI is.
- GPT-5 is available to all: free, plus, pro, and team/enterprise/EDU users.
There are two types of Openai models in this world: GPT and Reasoning Models. East’s advantages, such as GPT-4O, are that they add speed and accuracy, while it takes more time to use and use more calculation power to give better answers to the models such as O3 and O4. The latest model of Openai, GPT-5, is believed to provide all users access to the best of both models.
Also: According to Gartner’s 2025 Hype Cycle Report, Gen AI disillusionment
(Disclosure: ZDNET’s original company Ziff Davis filed a case of April 2025 against Openai, alleging that it violates Ziff Davis copyright training and operating its AI system.)
On Thursday, Openai Finally unveiled long-awaited GPT-5The family of the company’s next generation of families, which it is yet capable as the fastest, the fastest, the most clever and the most competent. GPT-5 is an integrated system that connects a smart model for most questions and an intensive logic for difficult problems (GPT-5 thinking).
If you are wondering what it does and if you should consider it even when you try it, then keep reading.
What is GPT-5?
The major difference between the GPT-5 and other Openai models is the real router feature, which allows the GPT-5 to automatically understand which models to use on the basis of interaction, the complexity of the prompt, and more. The router is constantly trained on real signals to understand the best scenario in which a model is to be used. Once a user hits the use range, a mini version of each model leads.
Also: Can GPT-5 Apple Apple fix intelligence? We are about to find out
While GPT-4o was a highly capable model, it was not a logic model like O3 and O4; They were limited to paying customers. As a result, GPT-5 is a particularly big win for free users, which usually did not reach any argument models and was excluded from the more advanced model, by default.
The GPT-5 family of the model is made up of GPT-5, GPT-Mini, GPT-5-Nano and GPT-5 Pro. The nuances between these models will mostly be subjects that are related to developers or enterprises, when which models choose to buy from API.
However, for most consumers, what you have to know is that GPT -5 will also be selected for free users automatically, and when the boundaries reach, GPT -5 will be switched to mini, which is still capable but more light model. The GPT-5 Pro is available only to the Chatgpt Pro subscribers, which comes at a cost of $ 200 per month per month. It is the most advanced version of GPT-5, which means for the most challenging and complex tasks-the work that will not encounter the possibility of average user.
How does GPT-5 perform?
With each model release, the GPT-5 drop was accompanied by benchmark evaluation, in which it earned a state-of-the-art score in Mathematics (AIME 2025), coding (SWE-Bench verified), and multimodal undersanding (MMMU). Even it performed competitively in the final examination of humanity, a new benchmark with multi-modal questions in more than 100 subjects, such as mathematics, science and humanities, as seen below.
The company claims that it is the strongest coding model ever, which is capable of creating websites, apps and games with simple text signals. In particular, Openai stocks stated that it has shown improvement in the complex front and generation and debugged large repository.
Also: Google’s Juogle AI coding tool comes out of beta with severe upgrade – more free tasks
Before the model was released, I saw a live demo of the feature, in which the user created a fully functional web app with interactive elements such as flashcard, a quiz with the right and wrong answer and a game from a simple text prompt. The final product looked smooth. As someone recently has a webpage experience, it took me hours to style using JavaScript and CSS. GPT-5 takes Vibe coding to the next level.
Even if you are more than an average GPT-5 user, who appoints AI to write, you will still get these benefits. Openai claims that GPT-5 is the most competent writing colleague, which is capable of dealing better than tasks, including free poetry such as “structural ambiguity”. Even if you use chat to write, you should see improvement.
People have been rapidly dependent on Chatgate for health related questions, as it is the ability to handle medical jerm, which can often be scary and scare. Now the experience is optimized with GPT-5, removing concerns, asking questions, understanding results, asking the provider’s questions and preparing the options.
Also: Chatgpt can now talk to you – more personality and other upgrades beyond GPT -5
Remember that GPT-5 does not replace a medical professional. Openai said the model performed the highest performance on the healthbench, which was in a benchmark evaluation of the company published earlier this year. External benchmarks are not yet standardized for how AI performs in medical scenarios.
How does GPT-5 handle security?
One of the biggest reforms available in GPT-5 is that the model is more accurate than any previous logic model and, according to OpenAI, is less hallucinations. The company stated that the GPT-5 reactions are 45% lower, with a factual error than the GPT-4O, in which the web search is capable of unknown signals, and 80% less openaiAIII O3 is likely to have a factual error than in OpenAI O3.
This is a major victory, as the logic models go beyond the prediction of the traditional pattern and are invited to “think”, which leaves the room for error.
Openai added new publicly available benchmarks to be available for a long time, which includes longFacts and Factskor, with GPT -5 showed a significant decline in hallucinations with thinking. GPT-5 (with thinking) also communicates more honestly with the user, when a task is impossible or not, sharing. This is important because the AI models often offer a commendable-sounding answer instead of accepting it that they do not know, which can increase the circulation of incorrect information. For more on evaluation results, take a look System card,
Another brand-new security facility is called “safe perfection”, which still enables the chat to answer, it will usually refuse. Instead, it will respond, but within the safety boundaries defined by Openai, and when it cannot, give it a clear explanation.
Finally, while not perfectly a safety issue, the model is less smoothing, or confirms, and uses less unnecessary emoji.
How can you reach GPT-5?
All users
GPT-5 and GPT-5 mini are available for all plus, pro, team and free users today, while enterprise and EDU users will get access next week. However, subscribers still receive tier the allowances. For example, included in $ 20 per month membership, Chatgpt Plus has “significantly more usage” limitations than customers free users. Meanwhile, the Chatgpt Pro users have unlimited GPT-5 and have access to GPT-5 Pro, one more advanced version of the model is included in their $ 200 per month membership. Openai said that enterprises and EDU users would be given “generous boundaries”.
Also: Microsoft rolls GPT -5 in its Copilot Suite – here you will find it
The GPT-5 will be set as a default model for every other models, replacing all other models for certified users. However, the paid users will still have the option to select it under the model picker. Openai stated that free users may not be able to use full arguments for a few days so far, given that once free user access to the extent, they will be transferred to GPT-5 mini.
Developers
The release of the model is also helpful for developers, as they can benefit from increased reliability and accuracy. To accommodate this, Openai is providing GPT-5, GPT-Min and GPT-5-Nano available in API. Two new parameters, arguments and verbosity are also to help develop developers that they need exactly the same from their model without overspeeding. Pro, plus, and team’s users can sign in Chatgpt to code with GPT -5 in Codex CLI.
The argument parameter makes GPT-5 cheaper to the tasks that do not require deep thinking, and then the Verbocity Parameter allows developers to do well that they want Verboz to want GPT-5. Pricing is cheaper than GPT-4O. For more information, you can Check the blog post,
What if I am not watching GPT-5 yet?
GPT-5 started rolling all plus, pro, team and free users when launched on Thursday. However, the rollouts are gradual, so if you checked immediately and did not see it, it is worth checking again to see if you have now. This was not shown to me till last night on my free, plus, or pro account.
Make sure you are signed in. Even if GPT-5 is available for free users, if you do not sign in, you will not have access to the latest features including GPT-5.
You can put my latest stories and technical adventures on social media. Follow me on Twitter/X @sabrinaa_ortiz And on Instagram @sabrinaa.ortiz,
