Breaking rest
OpenAI shuts down the Evals platform, including the Evals API, on November 30, 2026; it has been read-only since October 31, 2026, and OpenAI recommends migrating to Promptfoo
What changes
Notice date June 3, 2026. OpenAI's deprecations page lists the Evals dashboard, API, and graders for eval workflows with a November 30, 2026 shutdown date, status Read-only, and Migrate to Promptfoo as the recommended replacement; for fine-tuning graders it points to the fine-tuning timeline instead. The Evals guide says Evals became read-only for existing users on October 31, 2026, that existing evals content remains available during the transition window, and that the platform is scheduled to shut down on November 30, 2026. Read-only means calls that create or change evals or start runs should no longer be expected to succeed (the sources don't list which API calls already fail); from November 30 the platform, API included, shuts down, so reading evals, runs and results stops too. OpenAI's Promptfoo guide covers recreating an eval by hand in a promptfooconfig.yaml file; it does not require an OpenAI Evals export feature, and Promptfoo runs are separate from earlier OpenAI Evals runs.
Breaking: Calls fail after this date.
What to do
Recreate the evals you still need in Promptfoo (a promptfooconfig.yaml with your prompts, provider, test cases and assertions), following OpenAI's Promptfoo migration guide, and remove the OpenAI Evals API calls. Save any eval results you want to keep before November 30, 2026, and re-validate recreated graders, especially LLM-as-a-judge ones, before relying on their scores.
Strings to search for
/v1/evalsevals.createevals.retrieveevals.updateevals.listevals.deleteevals.runs