OpenAI is addressing some early bumps with GPT-5 by rolling out updates that boost the model’s capacity, quality, and user interface. CEO Sam Altman noted on X that ChatGPT Plus subscribers will see their GPT-5 rate limits double once the rollout is complete.
This update means you can still use the older “4o” model while OpenAI monitors usage to decide its continued availability. Currently, Plus users can send up to 100 messages every three hours, with an automatic switch to a smaller model if you go over the limit until it resets.
Additionally, the GPT-5-Thinking variant is capped at 200 messages per week, though switching between GPT-5 and GPT-5-Thinking doesn’t count against that limit. Altman promises you’ll notice an immediate improvement, saying GPT-5 is set to “seem smarter from today.”
A core feature of these updates is an automatic model switcher that chooses the best GPT-5 variant for each prompt. Initially, this system misfired for parts of the day, leading to a less-than-ideal performance. With API traffic having doubled over the past 24 hours, Altman describes the change as significant, even if the start was a bit bumpy.
Looking ahead, OpenAI is refining its strategy for model selection to make the experience more reliable, and they’re planning to show users which model is in use. Soon, you might even be able to manually toggle the “Thinking” mode from the interface, though we don’t yet have a timeline for this feature. For those who’ve experienced the frustrations of rigid automatic routing, these tweaks could make a noticeable difference.
