Codex Desktop Adds Free-to-Generate Next-Message Predictions for Pro Beta Users

Codex Desktop introduced Composer Predictions.

The feature uses the current Codex thread context to suggest a possible complete next message.

After Codex finishes responding, a suggestion may appear in the composer. Users can press Tab to accept it, edit it, and send manually. Accepting a prediction does not send the message automatically.

During beta, the feature is available for: - Personal ChatGPT Pro users; - Users aged 18 and older; - All supported regions; - The latest Codex desktop app; - Local and SSH threads; - GPT-6 Astra and GPT-6.1 Sol models.

A prediction will not appear after every response.

Predictions are enabled by default and can be disabled in Settings > General > Composer > Show predictions.

Predictions use the current thread context and do not independently retrieve memories or connected app information.

During the beta, generating predictions does not count toward Codex usage limits or consume credits. Messages users send, including accepted predictions, follow normal Codex usage limits and billing.

OpenAI has not announced a beta end date or post-beta pricing model.