Podcast Summary: Joe Rogan Experience for AI

Episode: New Updates to GPT-5: Inside the Next Evolution of AI
Date: August 18, 2025
Host: Joe Rogan Experience for AI

Overview

This episode dives deep into the latest updates to OpenAI’s GPT-5, focusing on the long-requested return of the model picker, the introduction of new model modes (Auto, Fast, Thinking, Pro), legacy model access, and the evolving conversation about AI personality customization. The host unpacks the significance of these changes—how they impact users, the motivations behind them, and what it means for the future of AI user experience.

Key Discussion Points and Insights

1. The Return and Redesign of the Model Picker

• [00:00–02:10]
  • OpenAI brought back the model picker, a feature users had missed due to confusion when it was removed.
  • Users can now choose between different response modes within GPT-5 (Auto, Fast, Thinking, Pro) rather than dealing with a complicated list of model names.
  • The “router” (which picks models for you) has been fixed, resolving previous issues of users receiving subpar results.

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

• [02:10–04:12]
  • Auto: The standard routing option, now supposedly reliable for general users.
  • Fast: For rapid responses.
  • Thinking: For deeper, more complex queries—users complained about “quick answers” lacking depth; this addresses that.
  • Pro: Tiered at $200/month, offering premium features like higher-end models, higher rate limits, and first access to experimental tools.

“Pro is actually an upgraded mode, I believe. If you want to get access to Pro, you have to pay $200 a month.” (Host, 03:00)

3. Return of Legacy Models and Model Tone

• [04:12–07:11]
  • Legacy models like GPT-4.0 have returned due to user demand, especially those who appreciated its “warmer tone.”
  • Users who rely on AI for more empathetic or therapeutic conversations preferred GPT-4.0.
  • OpenAI promises to give advanced notice before deprecating popular models in the future.

“People that use AI for like therapy said…they wanted [GPT-4.0] more. Sam Altman brought it back and said if they're ever going to depreciate it again…they’d give people lots of advance notice.” (Host, 06:00)

4. Technical Details: Rate Limits and Context Windows

• [07:11–08:40]
  • 3,000 messages per week limit for GPT-5 “Thinking” mode; after reaching the cap, users switch to “Thinking Mini.”
  • GPT-5 “Thinking” offers a 196,000-token context window—significant, though not the industry leader.
  • “Show additional models” toggle for paid users reveals even more legacy options (O3, 4.0, 4.1, 5 Thinking Mini).

5. Subscription and Paywall Dynamics

• [08:40–11:00]
  • The much-loved GPT-4.5 is now only accessible to Pro users due to high GPU/compute costs.
  • This exclusivity frustrates users who feel GPT-4.5 may outperform GPT-5 for their use cases.

“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 GPT-5 altogether.” (Host, 10:15)

6. Evolving AI Personalities: Warmth vs. Professionalism

• [11:00–14:00]
  • Sam Altman (via tweet) announced work on an updated GPT-5 personality aimed to be "warmer" but less “annoying” than the earlier GPT-4.0.
  • Growing recognition that users want customizable AI personalities, not just one-size-fits-all.
  • The host shares a story about a Ukrainian user dependent on GPT-4.0’s warmth, illustrating the real-world emotional impacts when model personalities shift.

“One learning for us from the past few days is…we really just need to get a world with more per-user customization of model personality.” — (Sam Altman, paraphrased by host at 12:30)

7. Looking Forward: The Importance of Customization

• [14:00–16:00]
  • Prediction: Users will soon personalize their AIs’ personalities, potentially receiving different answers from the same model depending on the customization.
  • The host cautions not to get too attached to legacy models, as they’re likely to be deprecated over time.
  • Highlights similar experiences across the industry (Anthropic’s Claude) where new models were less preferred for certain tasks (e.g., code-writing), validating OpenAI’s current path of restoring choice.

Notable Quotes & Memorable Moments

• Model Picker Frustration & Relief:
  “[The previous model picker] was so convoluted and confusing…They kind of consolidated it all and just said, okay, it’s all ChatGPT5 and we’ll pick which one you can use…The problem was [the router] was broken.” (Host, 01:10)

• Pro Subscription Skepticism:
  “200 a month. If you want access to the Pro…not sure I want to do for that model.” (Host, 03:40)

• AI as a Therapist:
  “A lady in Ukraine…talked to GPT4O and it was like her friend and therapist, whatever, and helped her out. And then when it got taken away…she was like super sad and depressed…” (Host, 12:00)

• Custom Personality Demand:
  “Should OpenAI be the one that makes the decision on like tuning the AI model’s warmth or its professionalism…? I, for one, would rather just pick what personality I want.” (Host, 13:20)

Timestamps for Important Segments

• [00:00] – Introduction; model picker saga
• [02:10] – Explanation of new GPT-5 modes
• [04:12] – Return and reasoning for legacy models like GPT-4.0
• [07:11] – Rate limits, context window, and “extra capacity”
• [08:40] – GPT-4.5’s status, GPU constraints, Pro subscription
• [11:00] – Sam Altman’s personality update plans, user stories
• [14:00] – Customization as the future of AI interaction
• [16:00] – Reflection on industry trends, closing thoughts

Takeaway

This episode offers a deep dive into the continuously evolving world of OpenAI’s GPT models, focusing on user autonomy, emotional resonance, and the technical/pricing structures that affect how people interact with AI. The host stresses a future where users can fully customize not only the technical capabilities of their AI tools, but also their personalities, providing a glimpse into a more individualized, responsive AI landscape.