GPT-5.4 vs GPT-5.4 Cyber

This comparison never had two sides, and now it has an ending. GPT-5.4 is a fully priced, fully specified model. GPT-5.4 Cyber was a name on a list — present in the pricing page's raw model enumeration, absent from the rendered price table, its own reference page returning nothing — and OpenAI removed it from the API at the start of October.

What the record shows

There was never a capability or price comparison to draw, because one side was never published in a checkable form. This page existed to say that plainly rather than let a reader searching for the pairing assume Cyber inherited GPT-5.4's numbers because they shared a generation name. That caution held to the end: the model was withdrawn without a price ever appearing.

What replaced it

OpenAI's deprecation notice points to "the most capable cyber model available to you." In practice that means the Daybreak program: Daybreak Blue gives approved defenders general-purpose models such as GPT-5.6 Sol, and Daybreak Red gives separately approved access to GPT-5.6 Cyber, which — unlike its predecessor — has a published price and a full model page. Both need approval before they can be called, so neither is offered in any calculator here.

What to use today

For ordinary work, GPT-5.4 remains a real, priceable API model, though no longer a Codex option for ChatGPT sign-in. For authorized security work, apply to Daybreak rather than looking for a successor id in the public price list.