OpenAI restores GPT-4o for paid ChatGPT users, vows “plenty of notice” if it deprecates 4o, and raises GPT-5 Thinking rate limits to 3,000 messages per week
OpenAI is once again making GPT-4o — the large language model (LLM) that powered ChatGPT before last week's launch of GPT-5 …
Context & Ripple Effects
OpenAI had already established a pattern of treating ChatGPT model availability as adjustable: it rolled back a GPT-4o update after concerns about sycophancy, while its earlier GPT-4 retirement plan kept that model available through the API. The immediate predecessor was a plan to let some users retain older-model selection only temporarily.
Restoring GPT-4o for paid users turns a planned transition into a negotiated one. The higher GPT-5 Thinking allowance also follows OpenAI’s stated response to rising reasoning-model use, making access policy and capacity allocation central to the paid product.
First-order effects
- Paid ChatGPT subscribers regain GPT-4o access and receive a clearer expectation of advance notice before any future removal, after a temporary old-model access plan had signaled a shorter runway.
- GPT-5 Thinking users can make up to 3,000 messages a week, expanding immediate access to the reasoning tier for workloads that had been constrained by limits.
Second-order effects
- Users and teams can defer migration from GPT-4o-based workflows, reducing the near-term pressure to re-prompt, re-evaluate, or adapt to GPT-5 behavior.
- OpenAI must operate both a legacy model and expanded reasoning access for longer, sharpening the trade-off between subscriber choice and increased reasoning-model availability.
Third-order effects
- If repeated, reversals around model retirement will make backward compatibility, transition notice, and model choice more important competitive features in hosted AI products.
- Usage limits are likely to remain a primary mechanism for segmenting access to compute-intensive models, rather than a one-time feature distinction between model generations.
The trend: Frontier AI subscriptions are evolving from simple model upgrades toward governed, capacity-aware access portfolios in which customers expect both choice and predictable migration paths.