After months of waiting, ChatGPT 5 has arrived in Israel and is now available to all users through the official website and app. The model, considered a technological leap compared to its predecessors, is already generating significant curiosity alongside reports of bugs and glitches. Here’s everything you need to know, from upgrades and new capabilities to the differences between GPT models.
ChatGPT 5, OpenAI’s new model, is presented as a significant step forward in the world of artificial intelligence. The company’s CEO, Sam Altman, described interacting with the system as "meeting a legitimate PhD-level expert." The company emphasizes that the new model is less prone to "hallucinations"—the phenomenon where the system provides unfounded or incorrect answers compared to previous versions.
The most notable innovation is that ChatGPT 5 is now the only model available for use. This means users will no longer be able to choose GPT-4o or any other version. Nevertheless, OpenAI claims there’s no need for concern, as the new system can automatically adapt to provide quick answers to simple questions or invest time and deep analysis into complex inquiries, depending on the user’s style of interaction.
Although all users externally receive ChatGPT 5, there are several internal operating modes, differing in depth of thought, processing time, and the amount of resources the system invests in an answer.
ChatGPT 5 Standard: This is the default model used for all queries. It balances speed and depth, providing answers within seconds with generally good accuracy. It is particularly suited for searches, daily questions, general writing, translation, message drafting, and idea generation.
ChatGPT 5 Fast: ("Fast") This mode is automatically activated when the system detects a simple or direct query. It minimizes response time but may rely on a more superficial analysis. This mode is ideal for generating concise and immediate answers but is less suitable for complex topics requiring in-depth review.
ChatGPT 5 Thinking: This is the most in-depth processing mode, currently available only to +Plus or Team users. Here, the system invests more time and resources, performing a multi-step "behind-the-scenes" process before returning the answer. During this process, it considers different options, checks reasoning, and sometimes even creates internal drafts before finalizing the response. The result is a longer, more substantiated answer, particularly suited for research, technical explanations, medical analyses, long-form creative writing, or complex software development.
ChatGPT 5 Light: When free users exceed their usage quota, the system automatically switches to lighter versions of ChatGPT 5, which use fewer resources. Here, quality may decrease slightly, especially in handling complex topics, but speed remains high.
The new system offers a significant upgrade in writing capabilities, with "literary depth and pacing" and features that allow translating complex texts—such as medical reports—into simple, clear language. In coding, the model has become more precise and creative, capable of adding "stylistic flair" to applications, not just dry functionality.
Another advantage is integration with Google services, including Gmail calendar and contacts. For example, ChatGPT 5 can create a personalized schedule based on calendar events and flight information from emails. However, not all users may want to share personal data with AI connected to third-party services due to privacy concerns.
Additionally, ChatGPT 5 significantly reduces "flattery" compared to previous versions. Internal tests found that the percentage of flattering responses dropped from 14.5% to less than 6%.
The global launch of ChatGPT 5 last weekend was accompanied by a flood of questions and criticism from users. During a Q&A session on Reddit, Sam Altman and the development team were repeatedly asked why GPT-5 felt "less smart" than the previous GPT-4o model and were even asked to restore the old model. Altman explained that the reason was a malfunction in the "router," a system component that decides in real time whether to use fast or deep processing. The malfunction caused GPT-5 to operate more superficially than usual for some time.
Altman promised that starting today, the system would operate more intelligently, and the company would make adjustments to improve model selection. He also mentioned that Plus users would still be able to use GPT-4o subject to further review. In addition, he committed to doubling the monthly usage limit for Plus subscribers during the rollout to allow them to get accustomed to the new model.
The launch was also accompanied by a controversy dubbed the "Graph Crime," when an incorrect graph was presented during the official presentation, showing GPT-5’s performance bar much higher even though the numerical result was lower. The image went viral as a joke on social media, and Altman admitted on Twitter that it was a "huge graph error."