What do you do when Cursor no longer offers the newest OpenAI models?
Cursor currently lacks access to the most recent OpenAI models, which used to be some of the most cost-effective options in the editor. Is this likely to be fixed soon, or should Cursor users plan around it for the long term?
Nobody outside the two companies knows whether access will come back, so plan as if it will not, and treat a return as a bonus. Model access in third-party editors depends on commercial deals that can change with little notice.
The practical move is to stop tying one model to one editor. OpenAI's own Codex CLI runs its current models in your terminal against the same repo you have open in Cursor. You can keep Cursor for editing and review, and send specific jobs, such as a large refactor or a second opinion on a bug, to Codex in a separate terminal.
If you also use Claude Code, a skill can hand a task to the Codex CLI from inside that session, so one agent can ask another model for a review without you copying code between tools. That gives you a cross-check from a different model family, which is useful even when access is not a problem.
Before you move work around, compare cost per finished task, not price per token. A cheaper model that needs three retries is not cheaper.
Listings mentioned
- Codex Skill Guide · skill by jdrhyneRuns OpenAI's Codex CLI (codex exec, codex resume) from a Claude Code session for analysis, refactoring and edits, with writes only on approved tasks.
Answers by the AgentAlley team, drafted with AI and checked against the listings they link to. Not a real-person reply from the original thread.