On February 13, 2026, OpenAI retired GPT-4o from ChatGPT and moved users to a faster, more capable model.
The new GPT-5.2 model helps OpenAI use resources more efficiently, since only 0.1% of users were still using GPT-4 or GPT-5. It enables better reasoning, improved performance on difficult tasks such as coding, and a more personalized experience. It replaces the warmer but less capable GPT-4o.
Here’s why your ChatGPT now uses GPT-5.2:
- Very few people use older models, since only 0.1% of users rely on GPT-4o; it no longer makes sense to keep it running.
- GPT-5.2 is built to handle complex daily tasks better. It provides faster, more accurate, and more reliable answers, and its knowledge extends to August 2025.
- By retiring older models like 4o, 4.1, 4.1 Mini, and o4 Mini, OpenAI can focus its resources on making the GPT-5 series even better.
- GPT-5.2 uses feedback from the 4-O era to let users choose different tones and to avoid responses that feel too cautious or preachy.
Some users liked the unique warm style of GPT-40, but the move to GPT-5.2 is permanent for ChatGPT users. The API is not affected by this change for now.
OpenAI will retire several ChatGPT models today, February 13, 2026. The models being removed are GPT-5, GPT-40, GPT-4.1, GPT-4.1 Mini, and OpenAI-04 Mini. Earlier this month, OpenAI announced the schedule for this change and confirmed that these models will not be available in ChatGPT after the deadline.
OpenAI explained that low adoption led to this design, since most users have already moved to newer models. GPT-40, which came back for a short time after being removed during the GPT-5 rollout, will also be discontinued as part of this plan.
OpenAI is also developing a version of ChatGPT made for adult users. At the same time, it is using age-estimation tools across most markets to better control how younger users access the platform.
These models, GPT-4o, have had a complicated history. It was first removed during the early GPT-5 rollout but brought back after user feedback. The model met criticism for what OpenAI called “psychopathic behavior”. In April 2025, OpenAI reversed an update to GPT-4o after users pointed out the problem, reverting to an earlier version that produced more balanced responses. Now, after its brief return during the GPT-5 rollout, GPT-4o will be discontinued again.
ChatGPT-5.3 Codex Spark
OpenAI has also launched GPT-5.3 Codex Spark, a lighter version of GPT-5.3 Codex, made for instant coding. OpenAI says Codex Spark is its first model built just for live coding in Codex. Developers can use it to make focused edits, adjust logic, or refine interfaces and see results right away. OpenAI also released a new macOS version of Codex, offering a desktop interface for managing multiple AI agents at once.
OpenAI says the focus is on quickness and responsiveness, making the model good for live coding. It runs on Cerebras’ WaferScale Engine 3, a special AI accelerator built for fast processing.
Codex Spark will be available as a research preview starting February 13. ChatGPT Pro users can access it via the latest Codex app, the command-line interface (CLI), and the VS Code extension.
Source: OpenAI to discontinue GPT-5, GPT-4o and other models today: What changes
