
OpenAI is giving ChatGPT users more control over how they interact with GPT-5, introducing new speed modes, increasing message limits and restoring GPT-4o to the model picker for paid subscribers. CEO Sam Altman shared the changes in a X post yesterday evening, saying they aim to improve flexibility while keeping performance high.
New GPT-5 speed modes

You can now choose between Auto, Fast, and Thinking when using GPT-5. Auto will likely be the go-to option for most users, automatically balancing speed and depth, while Fast prioritizes quicker responses and Thinking spends more time on reasoning-intensive tasks.
OpenAI is also increasing the weekly cap for GPT-5 Thinking to 3,000 messages, with extra capacity available via GPT-5 Thinking mini after hitting that threshold. The Thinking model’s context limit is now a hefty 196,000 tokens, meaning it can remember and process much longer conversations or documents in one go. Altman noted these limits may be adjusted over time depending on usage.
Model picker improvements and GPT-4o’s return

Following user backlash over GPT-4o’s removal, it’s now back in the model picker for all paid accounts by default. A new “Show additional models” toggle in ChatGPT web settings lets subscribers access other options like o3, 4.1, and GPT-5 Thinking mini.
GPT-4.5 remains exclusive to Pro users due to its high GPU costs.
Altman also revealed that GPT-5’s personality will soon get an update to feel “warmer” without becoming as polarizing as GPT-4o’s tone. Longer term, OpenAI plans to offer more per-user customization so people can tailor the AI’s style to their own preferences.
Bottom line
These updates give ChatGPT users more choice over speed, depth and personality, marking a quick response to recent user feedback.
With model variety returning and customization on the horizon, OpenAI seems intent on keeping its power users happy while broadening appeal.
Follow Tom's Guide on Google News to get our up-to-date news, how-tos, and reviews in your feeds. Make sure to click the Follow button.