OpenAI appears to be pruning parts of its Codex lineup: in a post on X, Tibo Sottiaux stated that GPT-5.2 and GPT-5.3-Codex will be “sunset” in Codex on June 2 for users logged in with a ChatGPT account. He also mentioned that, on free plans, GPT-5.5 will become the “default frontier model” for building and working in Codex, while the older models will remain available through the API.
Sottiaux added that GPT-5.2, released in December, now accounts for “<1% of usage in production” and is considered “quite outdated.” That claim drew pushback from some users, including people who described GPT-5.3-Codex as a go-to model for coding and code review, and who argued that it tends to be more usage-friendly than GPT-5.5 for longer sessions.
One reply asked whether the change included Codex Spark. Sottiaux responded that GPT-5.3-Codex-Spark is “not sunsetting,” but the clarification did little to calm concerns about the broader move. Several commenters complained that removing GPT-5.3-Codex would remove a lower-cost option they rely on when limits are already running low.
Others focused on the split between free and paid access. Some users pointed out that free plans getting GPT-5.5 by default appears to move in the opposite direction of what they expected from a rollout like this. A few commenters also argued that collapsing older Codex paths into newer ones may simplify OpenAI’s operations, but could remove fallback options if GPT-5.5 proves less stable for a given workflow.
The reaction suggests that, even if OpenAI views the older models as legacy infrastructure, a subset of users still treats them as practical tools rather than historical versions. For now, at least according to Sottiaux’s post, the cutoff date is June 2, and the older models will continue to exist on the API.
Source: Tibo on X
