OpenAI unveiled the GPT-5 models less than a week ago, but the launch of the long-awaited ChatGPT upgrade hasn’t been as smooth as the AI company might have hoped. GPT-5 was met with plenty of criticism soon after users started experiencing the new AI models that replaced all previous ChatGPT versions on the web, desktop, and mobile apps.
OpenAI responded to the backlash almost immediately, acknowledging the criticism surrounding GPT-5’s personality, performance, and usage limits and promising fixes. Some of them were immediate, as OpenAI deployed them over the weekend. The company brought legacy model access (GPT-4o) to the ChatGPT Plus tier. That’s the cheapest premium ChatGPT subscription, which wasn’t supposed to keep using the older models after the GPT-5 rollout. Also, Sam Altman announced higher limits for ChatGPT Plus users.
Updates to ChatGPT:
You can now choose between “Auto”, “Fast”, and “Thinking” for GPT-5. Most users will want Auto, but the additional control will be useful for some people.
Rate limits are now 3,000 messages/week with GPT-5 Thinking, and then extra capacity on GPT-5 Thinking…
— Sam Altman (@sama) August 13, 2025
Altman on Tuesday revealed more changes to ChatGPT that should improve the overall experience for all paid users. OpenAI brought back even more legacy models, so Plus users can switch back to their preferred AI chatbot versions rather than keep talking to the new models. On top of that, GPT-5 was split into multiple versions to make it easier for ChatGPT users to select the type of reasoning experience they want from the AI. Altman also said that OpenAI is working on changing GPT-5’s personality. The upcoming GPT-5 update “should feel warmer than the current personality but not as annoying (to most users) as GPT-4o.”
More legacy models for Plus users
Altman’s tweet on Tuesday addressed legacy ChatGPT models, announcing a big update. “4o is back in the model picker for all paid users by default. If we ever do deprecate it, we will give plenty of notice,” Altman said. “Paid users also now have a ‘Show additional models’ toggle in ChatGPT web settings which will add models like o3, 4.1, and GPT-5 Thinking mini. 4.5 is only available to Pro users—it costs a lot of GPUs.”
I can confirm that the “Show legacy models” setting changed to “Show additional models.” Since I had the toggle enabled, I now have access to GPT-4.1, o3, and o4-mini, in addition to GPT-4o (see the following screenshot). GPT-4o was enabled for ChatGPT Plus users a few days ago. ChatGPT Pro, Team, and Enterprise users had access to legacy models from the start, so they won’t have to make any changes to their ChatGPT experience.
In a separate tweet, Altman said OpenAI will prioritize compute for GPT-5 following the increased demand for the new models. First, paying ChatGPT users will get more total usage than before the GPT-5 rollout. Improvements to API demand will follow for existing customers. The quality of the free tier of ChatGPT will be improved after that. ChatGPT Free users did not receive any GPT-5 limit increases following Tuesday’s updates. Also, the Free accounts can’t use any legacy models.
New versions of GPT-5
“You can now choose between ‘Auto,’ ‘Fast,’ and ‘Thinking’ for GPT-5. Most users will want Auto, but the additional control will be useful for some people,” Altman said on X of the new GPT-5 experience. I’m a ChatGPT Plus user in Europe, and I can confirm that OpenAI’s ChatGPT user interface changes already went live here. Instead of the GPT-5 and GPT-5 Thinking models that showed up in the model picker before Altman’s tweet, I now have access to the GPT-5 versions above. Here’s how they’re described in the UI:
- Auto: Decides how long to think
- Fast: Instant answers
- Thinking mini: Thinks quickly
- Thinking: Thinks longer for better answers
This is a significant change to the GPT-5 experience, signaling that OpenAI’s vision of deploying a single model that can decide on its own when to reason and when to answer quickly isn’t working as intended.
The ChatGPT UI also includes a GPT-5 Pro model which is grayed out for ChatGPT Plus users. Altman said a few days ago that OpenAI plans to give Plus users access to a few GPT-5 Pro conversations per month. These are only available in higher-priced ChatGPT subscriptions. As of Tuesday, the rollout of GPT-5 Pro to Plus users remains uncertain. “We do not have the compute to do it right now; we would like to at some point!,” Altman said in a reply to a user’s question on X.
GPT-5 limits for ChatGPT Plus users
Originally, ChatGPT Plus users were able to send 80 GPT-5 messages every three hours and 200 GPT-5 Thinking messages per week. GPT-5 auto-switching to Thinking mode would not count towards that 200 message limit. On Sunday, Altman said GPT-5 Thinking access will go up to 3,000 messages a week for ChatGPT Plus users. GPT-5 access will double to 160 messages every three hours.
OpenAI updated its GPT-5 support document to mention the updated limits. Since ChatGPT Plus and Team users have unlimited access to GPT-5 models, the limits in Altman’s tweet apply (mostly) to the ChatGPT Plus tier. There is a quirk for ChatGPT Team users. If they select GPT-5 Thinking manually from the model picker, then they’re also subject to the 3,000 messages per week limit.
As before, automatic switching between GPT-5 and GPT-5 Thinking will not count towards that 3,000 ceiling. Also, GPT-5 will resort to reasoning (thinking longer) even if that 3,000 limit is reached.
