OpenAI unveiled its newest Instant model, GPT‑5.5 Instant, in a brief system card posted to the company blog. The release follows the GPT‑5.3 Instant baseline, as there is no intervening GPT‑5.4 version. By naming the model gpt-5.5-instant, OpenAI separates it from the similarly labeled GPT‑5.5 Thinking variant, which serves a different use case, and will be available to enterprise customers starting next month.

OpenAI treats GPT‑5.5 Instant as a High capability system within its Cybersecurity and Biological & Chemical Preparedness categories, triggering stricter safety mitigations than prior Instant models. The mitigation framework mirrors earlier releases, but the elevated classification means monitoring, usage restrictions, and protocols, and includes real‑time audit logging for all outputs. This move signals the model’s expanded potential for sensitive threat‑analysis tasks while preserving alignment safeguards.

Developers can invoke the model via the identifier gpt-5.5-instant, ensuring API calls target the correct version. By clarifying the distinction from GPT‑5.5 Thinking, OpenAI reduces confusion for users building security‑oriented applications. The announcement underscores the company’s incremental approach: each Instant iteration builds on the last while tightening controls where the model’s power meets higher‑risk domains. It also supports fine‑tuning for specialized threat models.