GPT-5 is going so well for OpenAI that there's now a 'show additional models' switch
(2025/08/13)
- Reference: 1755102610
- News link: https://www.theregister.co.uk/2025/08/13/gpt5_updated_again/
- Source link:
There has been more furious backpedalling from OpenAI following the company's ill-judged launch of GPT-5 and the removal of previous model selection.
El Reg's essential guide to deploying LLMs in production [1]READ MORE
The plan for GPT-5 was that one model would rule them all, so to speak, with GPT-5 directing user requests to the appropriate destinations. But it's fair to say that things have not [2]gone according to plan since the technology's [3]debut last week.
After a weekend of dealing with user protests, OpenAI CEO Sam Altman eventually [4]confirmed the 4o model was "back." OpenAI has since gone further in its attempts to respond to feedback, while also straying away from the goal of GPT-5 being a unified model that dispensed with the need for fiddling with settings and model selection.
"You can now choose between 'Auto', 'Fast', and 'Thinking' for GPT-5," Altman [5]wrote this morning. "Most users will want Auto, but the additional control will be useful for some people."
Presumably he's talking about those people who objected to GPT-5 making the decision for them.
[6]
Rate limits are now 3,000 messages per week with GPT-5 Thinking.
[7]
[8]
Altman reiterated that "4o is back in the model picker for all paid users by default," and said "plenty of notice" will be given should it ever be deprecated. Older models can be manually selected too. Altman added, "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."
The updates are an admission that GPT-5's routing of requests is not to the taste of every user, and it needs to do more work. After Altman's admission on X, OpenAI VP and Head of ChatGPT, Nick Turley, [9]admitted , "We're not always going to get everything on try #1," but said he was "very proud of how quickly the team can iterate."
[10]OpenAI pulls plug on ChatGPT smarmbot that praised user for ditching psychiatric meds
[11]Meta bets you want a sprinkle of social in your chatbot
[12]AI bigwigs urge AGs to block OpenAI's profit pivot
[13]As ChatGPT scores B- in engineering, professors scramble to update courses
OpenAI's GPT-5 problems have not stopped its adoption. Yesterday, Microsoft [14]confirmed the model was now rolling out in GitHub Copilot for Visual Studio and described it as "OpenAI's most advanced model to date."
'Suddenly deprecating old models' users depended on a 'mistake,' admits OpenAI's Altman [15]READ MORE
Altman concluded by saying that OpenAI was "working on an update to GPT-5’s personality which should feel warmer than the current personality but not as annoying (to most users) as GPT-4o.
"One learning for us from the past few days is we really just need to get to a world with more per-user customization of model personality." ®
Get our [16]Tech Resources
[1] https://www.theregister.com/2025/04/22/llm_production_guide/
[2] https://www.theregister.com/2025/08/11/openai_tweaks_gpt5_after_user/
[3] https://www.theregister.com/2025/08/07/openai_gpt_5/
[4] https://www.theregister.com/2025/08/11/openai_tweaks_gpt5_after_user/
[5] https://x.com/sama/status/1955438916645130740
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aJ0K9kQhL9a1kkOpVVZAIwAAAAQ&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aJ0K9kQhL9a1kkOpVVZAIwAAAAQ&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aJ0K9kQhL9a1kkOpVVZAIwAAAAQ&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[9] https://x.com/nickaturley/status/1955441658050515144
[10] https://www.theregister.com/2025/04/30/openai_pulls_plug_on_chatgpt/
[11] https://www.theregister.com/2025/04/29/metas_standalone_ai_app/
[12] https://www.theregister.com/2025/04/23/openai_for_profit_stop_urged/
[13] https://www.theregister.com/2025/04/23/whats_worth_teaching_when_ai/
[14] https://devblogs.microsoft.com/visualstudio/gpt-5-now-available-in-visual-studio/
[15] https://www.theregister.com/2025/08/11/openai_tweaks_gpt5_after_user/
[16] https://whitepapers.theregister.com/
El Reg's essential guide to deploying LLMs in production [1]READ MORE
The plan for GPT-5 was that one model would rule them all, so to speak, with GPT-5 directing user requests to the appropriate destinations. But it's fair to say that things have not [2]gone according to plan since the technology's [3]debut last week.
After a weekend of dealing with user protests, OpenAI CEO Sam Altman eventually [4]confirmed the 4o model was "back." OpenAI has since gone further in its attempts to respond to feedback, while also straying away from the goal of GPT-5 being a unified model that dispensed with the need for fiddling with settings and model selection.
"You can now choose between 'Auto', 'Fast', and 'Thinking' for GPT-5," Altman [5]wrote this morning. "Most users will want Auto, but the additional control will be useful for some people."
Presumably he's talking about those people who objected to GPT-5 making the decision for them.
[6]
Rate limits are now 3,000 messages per week with GPT-5 Thinking.
[7]
[8]
Altman reiterated that "4o is back in the model picker for all paid users by default," and said "plenty of notice" will be given should it ever be deprecated. Older models can be manually selected too. Altman added, "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."
The updates are an admission that GPT-5's routing of requests is not to the taste of every user, and it needs to do more work. After Altman's admission on X, OpenAI VP and Head of ChatGPT, Nick Turley, [9]admitted , "We're not always going to get everything on try #1," but said he was "very proud of how quickly the team can iterate."
[10]OpenAI pulls plug on ChatGPT smarmbot that praised user for ditching psychiatric meds
[11]Meta bets you want a sprinkle of social in your chatbot
[12]AI bigwigs urge AGs to block OpenAI's profit pivot
[13]As ChatGPT scores B- in engineering, professors scramble to update courses
OpenAI's GPT-5 problems have not stopped its adoption. Yesterday, Microsoft [14]confirmed the model was now rolling out in GitHub Copilot for Visual Studio and described it as "OpenAI's most advanced model to date."
'Suddenly deprecating old models' users depended on a 'mistake,' admits OpenAI's Altman [15]READ MORE
Altman concluded by saying that OpenAI was "working on an update to GPT-5’s personality which should feel warmer than the current personality but not as annoying (to most users) as GPT-4o.
"One learning for us from the past few days is we really just need to get to a world with more per-user customization of model personality." ®
Get our [16]Tech Resources
[1] https://www.theregister.com/2025/04/22/llm_production_guide/
[2] https://www.theregister.com/2025/08/11/openai_tweaks_gpt5_after_user/
[3] https://www.theregister.com/2025/08/07/openai_gpt_5/
[4] https://www.theregister.com/2025/08/11/openai_tweaks_gpt5_after_user/
[5] https://x.com/sama/status/1955438916645130740
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aJ0K9kQhL9a1kkOpVVZAIwAAAAQ&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aJ0K9kQhL9a1kkOpVVZAIwAAAAQ&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aJ0K9kQhL9a1kkOpVVZAIwAAAAQ&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[9] https://x.com/nickaturley/status/1955441658050515144
[10] https://www.theregister.com/2025/04/30/openai_pulls_plug_on_chatgpt/
[11] https://www.theregister.com/2025/04/29/metas_standalone_ai_app/
[12] https://www.theregister.com/2025/04/23/openai_for_profit_stop_urged/
[13] https://www.theregister.com/2025/04/23/whats_worth_teaching_when_ai/
[14] https://devblogs.microsoft.com/visualstudio/gpt-5-now-available-in-visual-studio/
[15] https://www.theregister.com/2025/08/11/openai_tweaks_gpt5_after_user/
[16] https://whitepapers.theregister.com/
Anonymous Coward
What do you expect. He's not the sharpest spoon in the drawer.
"One learning for us from the past few days is we really just..."
The word is "lesson"!