Podcast Summary: The AI Podcast

Episode: New Updates to GPT-5: Exploring Cutting-Edge Features
Date: August 18, 2025
Host: The AI Podcast

Overview

In this episode, The AI Podcast host provides an in-depth analysis of the latest updates to OpenAI's GPT-5, including major changes to the model picker interface, the return of older models like GPT-4o, the introduction of selectable response modes, and new features that emphasize user control and customization. The discussion addresses OpenAI's response to user feedback, subscription tiers, model context and rate limits, and the larger trend toward personalizing AI personalities.

Key Discussion Points & Insights

1. Model Picker Revamp: More Control, Simpler Choices

[00:00 - 04:15]

• Return of Model Picker: OpenAI reinstated the model picker after significant user feedback. Older, previously deprecated models are now available again.
  • “OpenAI has officially brought back their model picker... They’ve added new features to GPT5, new options to choose how you get your responses...” (00:00)
• Reason for Change: Many users found the previous dropdown interface confusing due to the plethora of model options (e.g., O3 mini, 4.5, etc.).
• Model Routing: Initially, GPT-5 defaulted to an automatic router to select the best model per query. However, this router was unreliable, often delivering less accurate results, sparking user complaints.
  • “...when they rolled out GPT5, the router was broken. So you're just getting like worse results for most of the time. And people were saying GPT5 was worse than anything before.” (01:34)

2. New Response Modes: Auto, Fast, Thinking, Pro

[04:15 - 09:00]

• Modes Explained:
  • Auto: The default; OpenAI's improved router chooses the best mode. Recommended for most users.
  • Fast: For quick answers.
  • Thinking: For more in-depth, thoughtful answers to complex queries.
  • Pro: Premium tier, bundled with highest performance capabilities and exclusive tools (e.g., Sora video generation, Operator agents).
• Pro Tier Access: Costs $200/month. Offers higher quality responses and access to exclusive or experimental features.
  • “Pro is actually an upgraded mode, I believe. If you want to get access to Pro, you have to pay $200 a month... there's a bunch of extra things you get with Pro, and it's kind of this revolving list of their newest products...” (02:39)

3. The Return and Importance of GPT-4o

[09:00 - 12:30]

• User Demand: GPT-4o’s “warmer” conversational style was missed, especially by those using it for therapeutic or companionship purposes.
  • “A lot of people have requested this back basically because they said it had like a warmer tone. It was nicer to talk to... people that use AI for like therapy said they wanted this more.” (09:25)
• OpenAI's Commitment: Promised to give advance notice before deprecating popular models.
  • “Sam Altman brought it back and said if they're ever going to deprecate it again in the future, they give people lots of advance notice.” (09:45)

4. Technical Changes: Rate & Context Limits, Model Availability

[12:30 - 15:25]

• Message Limits: 3,000 messages per week with GPT-5 Thinking; after the cap, users are shifted to Thinking Mini.
  • “Rate limits are now 3000 messages a week with GPT5 Thinking and then extra capacity on GPT5 Thinking Mini after that limit. Honestly, 3000 messages a week for the thinking model I think is probably great for most people.” (12:57)
• Context Window: 196,000 tokens for GPT-5 Thinking; competitive, though not as large as the million-token windows seen elsewhere.
• Legacy Models Toggle: Paid users can activate additional models like O3, 4.0, 4.1, and 5 Thinking Mini from settings. GPT-4.5 remains Pro-only.

5. Subscription Tiers & Model Access

[15:25 - 18:00]

• Model Restrictions: GPT-4.5, favored for its performance, is locked behind the $200 Pro tier due to high computational demands.
• User Frustrations: Some express annoyance that 4.5, which they found superior to GPT-5 in some tasks, is now paywalled.
  • “...if they're basically paywalling 4.5 behind a $200 a month subscription, it makes me wonder if that one is not just a better model than GPT5 altogether. And now I'm kind of, you know, mad they took it away from me for allegedly a better model that's worse.” (17:09)

6. Personality and Customization: The Next Frontier

[18:00 - 23:30]

• Personality Tuning: OpenAI is working to make GPT-5's personality “warmer,” but less “annoying” than GPT-4o.
• Memorable Quote:

  “We are working on an update to GPT5's personality, which should feel warmer than the current personality, but not as annoying to most users as GPT4O. However, one learning for us... is that we really just need to get a world with more per user customization of model personality.”
  — Sam Altman, as quoted by the host (18:14)

• User Experiences: The story of a user in Ukraine who lost access to the more empathetic GPT-4o highlights the importance of emotional tone in AI.
  • “She lost a lot of family...her city was being bombed. She talked to GPT4O and it was like her friend and therapist...Then when it got taken away when GPT5 came out, she was super sad and depressed because GPT5 wasn't as like... warm or whatever.” (20:22)
• Broader Implications: The host argues that users, not OpenAI, should decide AI personality characteristics. The future lies in deep user customization to match individual preferences and communication needs.

7. Industry Context: Similar Challenges for Other AI Providers

[23:30 - End]

• Not Just OpenAI: Anthropics’ Claude faced similar backlash when its new model was less favored for certain tasks; coders preferred older versions for coding.
  • “Claude and anthropic actually went through a very similar thing. They came out with a brand new model and coders preferred the older model...” (23:44)
• Host’s Advice: Don’t get too attached to legacy models—AI development is iterative and discontinuations will continue.

Notable Quotes & Memorable Moments

• On Model Picker Confusion:

  “There’s so many different models. Very confusing. They kind of consolidated it all and just said, okay, it’s all Chat GPT5 and we’ll pick which one you can use.” (01:16)

• On Subscription Frustration:

  “...if they're basically paywalling 4.5 behind, behind a 200amonth subscription, it makes me wonder if that one is not just a better model than GPT5 altogether.” (17:09)

• On Personality Customization Future:

  “We really just need to get a world with more per user customization of model personality.” (18:25)

• On User Sentiment:

  “…a lady in the Ukraine…her city was being bombed. She talked to GPT4O and it was like her friend and therapist...then when it got taken away...she was super sad and depressed.” (20:22)

Important Timestamps

• 00:00 — Model picker returns, overview of episode focus
• 01:34 — GPT-5 router issues and user dissatisfaction
• 02:39 — Introduction of modes (Auto, Fast, Thinking, Pro) and pricing
• 09:25 — User demand for GPT-4o's warmer personality
• 12:57 — Clarification of rate and context limits
• 17:09 — Model 4.5 moved into $200 Pro tier, user frustrations
• 18:14 — Sam Altman's statement on personality updates and customization
• 20:22 — Story about the Ukrainian user and the emotional significance of AI tone
• 23:44 — Similar issues with other AI providers, e.g., Claude and Anthropic

Conclusion

This episode highlights OpenAI’s rapid and sometimes bumpy evolution of its offerings, balancing advanced new features with user demands for simplicity, warmth, access, and control. OpenAI's responsiveness to feedback—seen in the return of legacy models and continual tweaks—signals an industry trend toward deeper user agency: soon, not only model performance but also its tone and tactics may be entirely user-defined.