The announcements of the upcoming GPT-5 were heated by the media for about two weeks. Altman himself tried various comparisons (even to the Manhattan project) to prove that the new GPT-5 will be a revolution on the LLM market. Did it finally happen? There are actually a lot of news, which is why I spread them into prime factors!
GPT-5, or PHD in your pocket!
It cannot be denied that Chatgpt is currently one of the most popular large language models, i.e. AI chatbots. Openai and the startup CEO of Altman himself are well aware of this, so they constantly want to keep the hegemon position. It’s just that the competition is not sleeping and definitely begins to trample Openai on cybernetic heels. Just look at the capabilities of Gemini 2.5 Pro or the next Claude iteration or Groc from Elon Musk.
That is why for over a month Altman himself has prepared the land of public opinion for the fact that the Prime Minister ChatgPT-5 is approaching rapidly. According to Altman (who hosted this past weekend), the possibilities of the GPT-5 are so huge that at first they scared him.
On August 7, during the OpenAI Summer Update, presented GPT-5 of the world. In fact, there are a lot of new products, but the new model is another iteration of a well-functioning LLM rather than a actual revolution in the geni industry.
The biggest novelty (and at the same time the smartest marketing movement) is the disclosure of “full” GPT-5 free users. Yes, you read well. Opeli decided to be generous … in its own style, because of course with limits – up to 10 answers every five hours, after which the system transfers to the mini model. Deep Research functions are also very limited. It’s a bit like he gets a perfume samples instead of the whole bottle, but it is enough to feel the difference in quality.
News that you can’t pass by indifferently!
The context window increases to 400,000 tokens. Is that a lot? Very. You can upload about 600 pages in one prompt and get satisfactory results. This completely changes the way the GPT-5 can be used, because with such a large cones we get almost a small RAG. As part of the Wytmłuja – RAG (Retrieval -Agumented Generation, we call AI systems, combining search for information from specific databases with text generation.
It is worth adding that the GPT-5 is really more AI because it contains three models:
- GPT-5
- GPT-5 Mini
- GPT-5 NANO
What does this mean? Well, the GPT chooses which model will be necessary to make up with your prompt. This significantly reduces costs, which is particularly important using the GPT-5 by API. Opeli called this way of choosing models as a real-time router.
Better understanding of complex instructions
GPT-5 actually better times with multi-stage commands without breaking them into prime factors. What does it matter? Large, especially in business, because less iteration means lower costs. For users, this is also a fundamental difference, because the model understands the pleasure better, which allows for more complex tasks. According to developers from OpenAI, GPT-5 is to be distinguished by the fact that it is definitely hallucinating. This is of course a theory, because in practice users report a large amount of hallucinations. While writing this article, I came across at least two glaring hallucinations. This can of course be explained by the fact of the massive interest of GPT-5, but Fact-Checking still remains the most effective weapon against the fact that LLMs make up facts.
Small changes that matter
The new model from OpenAI has also introduced a lot of minor changes that matter. First of all, we can now set the personalization of the chat by default, i.e. influence in what character it answers our questions. It can be a substantive, funny or even ironic nature. Voice Mode, which sounds even more realistic when it comes to intonations, has similar amenities.
It is worth adding that the GPT-5 can now be integrated with Gmail and the Google calendar. An interesting development of security technologies is “Safe-Comlation”, i.e. a substantive refusal to complete the task. Example? You write Prompt and ask for something that violates OpenAI safety standards. From now on you will not only get a refusal, but a substantive explanation why a given prompt cannot be properly processed.
Programmers in the attention center
Here OpenAi decided to play Va Banque. If Altman’s start-up is to start earning real money (and not to live from subsidies and PR), then she must convince developers-those who are willing to pay thick thousands for tools that really improve their performance.
And I have to admit – at first glance it looks promising. GPT-5 better understands the intentions of the programmer from simpler prompts, creates more extensive applications with fewer errors. Earlier it was also possible, but now it is simply faster and easier – especially for those who know about coding as much as I did about Alpac breeding;) A very interesting example was applications that were created on a live demo during the OpenAi Summer Update. It was even an application for learning a French language with specific achievements, and even a simple educational game based on snake.
Of course, we have to wait for the final rating for professionals to spend more time with GPT-5 in real projects. But if OpenAI actually managed to significantly improve the quality of the generated code, it can be a game-changer in the fight against competition.
Iteration, not a revolution
The most important question is: GPT-5 is a revolution? The answer is simple – no. It’s evolution, and it’s quite thought out.
Opeli does not try to turn the market upside down. Instead, he hits where he lost his position, giving himself breath and time to prepare real “bombs”. Because the real revolution is to come with Aga and the Stargate project – these hundreds of billions of dollars worth data centers filled with millions of GPU, without which the mentioned Aga will not arise.
GPT-5 is to do something simpler, but equally important, i.e. stop users and let OpenAI calmly prepare the next move. This is a defensive strategy, but it does not mean weakness. This means that the company knows what it is doing.
