Overview
- GPT-4o and other legacy models are available again to ChatGPT Plus and Team subscribers via the “show legacy models” setting.
- Paid users now receive up to 3,000 weekly reasoning requests after limits were significantly raised to accommodate demand.
- OpenAI has started UI updates to show which model responds to each query and promised transparency on capacity tradeoffs.
- Early GPT-5 deployments drew criticism for slower responses, hallucinations and easy jailbreak exploits, spurring ongoing performance and safety patches.
- The push-and-pull over model choice underscores OpenAI’s challenge in balancing system efficiency, user loyalty and competitive pressures from Anthropic, Google and Meta.