Sam Altman Unveils Major Upgrades and Fresh Features for GPT-5 Amidst Criticism

Tech
By 24matins.uk,  published 13 August 2025 at 17h15, updated on 13 August 2025 at 17h16.
Tech

Sam Altman has addressed recent criticism surrounding GPT-5 by unveiling significant improvements and introducing new features. These updates aim to enhance user experience and performance, reflecting OpenAI’s ongoing commitment to advancing artificial intelligence capabilities.

Tl;dr

  • OpenAI adds new speed modes for GPT-5.
  • GPT-4o returns for all paying users.
  • Enhanced personalization and model selection features launched.

    • OpenAI Prioritizes Flexibility Amid User Demands

    As the landscape of artificial intelligence evolves rapidly, OpenAI finds itself under increasing scrutiny from a vocal and diverse user base. The company has responded in recent hours with a string of significant announcements, all centered around giving users more freedom in their experience with ChatGPT. In direct response to persistent — and sometimes critical — feedback, these changes underscore a clear shift: more choice, more control.

    GPT-5 Introduces Speed Modes and Expanded Limits

    Arguably the headline feature, the rollout of three distinct interaction speeds for GPT-5 marks a notable advance. Now, users can toggle between « Auto », « Fast » and « Thinking ». While « Auto » strikes a balance between speed and analysis by default, those requiring immediate responses may turn to « Fast ». Conversely, for tasks demanding more depth, « Thinking » delivers enhanced analytical capability — with a raised weekly cap now set at 3,000 messages. That figure can be further increased thanks to the new GPT-5 Thinking mini option.

    Another technical milestone worth highlighting is the context limit expansion: conversations or documents can now stretch up to 196,000 tokens, allowing far broader exchanges. According to CEO Sam Altman, these thresholds remain subject to adjustment as usage patterns emerge.

    Diversity Restored: The Return of GPT-4o and Expanded Model Selector

    One change drawing immediate attention is the restoration of the popular GPT-4o model. After considerable community advocacy for greater diversity in available models, all paying subscribers once again have access to this option by default. Accessing other variants — including o3, 4.1, or GPT-5 Thinking mini — is now as simple as pressing a button in the web interface settings. However, it’s important to note that GPT-4.5, due to high GPU processing costs, remains exclusive to Pro subscribers.

    In tandem with these updates, Sam Altman has revealed that an upcoming release will imbue GPT-5‘s conversational style with warmer personality traits — albeit carefully avoiding some divisive quirks observed previously in GPT-4o. The overarching aim? To let users tailor the AI’s tone more closely to their individual preferences over time.

    The Road Ahead: Customization at the Core

    To summarize the practical shifts introduced:

  • Select interaction mode (Auto, Fast, Thinking) per need.
  • A richer palette of models available for each user.
  • User-driven adjustments shaping future developments.

    • Ultimately, these moves signal a dual ambition from OpenAI: retaining its expert user base while simultaneously courting newcomers by enhancing flexibility and personal agency within ChatGPT. In today’s hypercompetitive AI sector, listening attentively — then acting decisively — may well become one of its greatest assets.