If you haven't heard, OpenAI's launch of ChatGPT-5 on August 7 has caused a furor among a contingent of paying ChatGPT subscribers who find the new frontier model's communication style too clinical compared to legacy model GPT-4o's warmer, more conversational approach. Keep reading to make the old model available to select again in the ChatGPT app for Mac.
After just a few days of trying GPT-5, many Plus subscribers have decided they prefer GPT-4o's personality and creative collaboration skills. As a result, OpenAI is allowing paying customers to continue to use the prior-generation GPT-4o model if they don't want to use the new GPT-5 model. If you count yourself among them, here's how to bring the older model back into your ChatGPT for Mac experience.
Re-enabling GPT-4o Access
To restore GPT-4o in your ChatGPT for Mac app, you first need to enable the setting through the web interface:
- Visit chatgpt.com in your web browser and log in to your account.
- Click your profile picture in the bottom-left corner, then choose Settings.
- Navigate to the General section.
- Toggle on the switch next to "Show legacy models."
The setting takes effect immediately in the web interface, where you'll find GPT-4o listed under "Legacy models" in the model selector dropdown.
Give it a few minutes, and the setting will automatically sync with ChatGPT for Mac upon restarting the app. You'll then find GPT-4o available in the same model selector within the desktop app. If you have an established workflow built around GPT-4o's specific response patterns and interaction style, it can now continue without disruption.
Note that GPT-4o legacy access is currently limited to paying ChatGPT users. The cheapest option is the Plus plan ($20/month). OpenAI hasn't specified how long legacy model support will continue, saying they'll "watch usage as we think about how long to offer legacy models."
Top Rated Comments
It sounds unreal, something straight out of a scifi movie like “Her” , but reality is already exceeding fantasy.
Honestly i don’t think Apple is making a mistake by taking their time with AI. Some of the most nefarious aspects of AI are just starting to show up.
ChatGPT spouts nonsense based on probabilities drawn from the pool of global knowledge.
Perplexity spouts nonsense based on current search results.
This means that the information provided by Perplexity can sometimes be more reliable. However, it can also take you in completely the wrong direction if there are name overlaps.
ChatGPT is more detailed, but it does not cite sources in the first step and often has to admit that it has fantasized when asked.
Now I get the best of both worlds
Someone made a screen shot:
Shouldn't the height of such a bar be proportional to the value it represents?