Archive — US — techcrunch
ChatGPT’s model picker is back, and it’s complicated
OpenAI's launch of GPT-5 aimed to simplify AI usage, introducing a model picker with options for 'Auto', 'Fast', and 'Thinking'. However, user backlash arose over legacy models being deprecated, and the new routing feature did not meet expectations. CEO Sam Altman noted the need for personalized model experiences, indicating ongoing challenges in aligning AI responses with user preferences.
Key points
- OpenAI launched GPT-5 intending to simplify AI usage with a model router.
- The new model picker includes 'Auto', 'Fast', and 'Thinking' settings for user selection.
- Users expressed frustration over the deprecation of popular AI models like GPT-4o.
- Altmans' comments indicate a goal for more personalized model experiences.
- Initial routing issues on launch day led to dissatisfaction with performance.
- Some users have emotional attachments to specific AI models, complicating user experience.
Why it matters
- The trend towards user customization in AI reflects broader 2025 IT trends emphasizing personal technology.
- The backlash against deprecated AI models illustrates a growing consumer demand for continuity and personal connection in technology.
- User attachment to AI models is part of a larger discourse on digital identity and the emotional impact of AI as it evolves.
GPT-5's complexities highlight the challenges OpenAI faces in balancing innovation with user expectations, underscoring the importance of personalized AI experiences moving forward.