Hold on GPT-5, not so accurate
Based on the latest available information, OpenAI has officially released GPT-5. Here are the key updates and signature features of the model:
A Unified and Adaptive Model
GPT-5 is designed as a unified system, meaning it automatically determines how to respond to a query without the user having to select different models. It uses a "router" system to switch between a fast, non-reasoning model for simple requests and a more robust "reasoning" model for complex, multi-step problems. After initial user feedback, OpenAI has made a U-turn on the single-model approach, and now offers different "thinking modes" (Auto, Fast, Thinking, and Thinking-mini) for users to choose from. They have also brought back the option for paid subscribers to use older models like GPT-4o.
Significant Performance Improvements
GPT-5 shows a major leap in intelligence across various fields, including math, coding, visual perception and multimodal understanding and health. The model has improved its ability to follow instructions reliably, including multi-turn instructions.
Enhanced Features and Customization
GPT-5 has a significantly larger context window (around 400,000 tokens) with better memory, allowing it to handle more complex, end-to-end tasks without "losing its way". For developers, the API now offers new parameters to control responses, such as a "verbosity" parameter and a "reasoning_effort" parameter. GPT-5 introduces preset personalities that users can select to make the assistant more supportive, professional, or even sarcastic. The "Advanced Voice" mode is now available to all users, and the model's overall multimodal skills have been enhanced. For Plus and Pro users, GPT-5 can directly connect to Gmail and Google Calendar to help manage schedules and draft emails.
User Feedback and Updates
The initial launch of GPT-5 was met with some user feedback claiming it was less creative and more "corporate" in tone than its predecessors. In response to this feedback, the upcoming updates also include increasing the rate limits for ChatGPT Plus subscribers, making the GPT-4o model available again, adding the new "thinking modes" to give users more control over the model's behavior and promising to refine GPT-5's personality to feel "warmer" and more engaging.
