OpenAI: GPT-5.5 Instant becomes the new default ChatGPT model with fewer hallucinations

OpenAI introduced GPT-5.5 Instant on May 5, 2026, as the new default model replacing the previous instant variant in ChatGPT. According to the brief official announcement, the model delivers smarter and more precise responses, a reduced hallucination rate, and better conversation personalization.

Simultaneously with the model announcement, OpenAI published an accompanying system card — a document describing the model’s technical characteristics, safety evaluations, and limitations.

What does GPT-5.5 Instant change in ChatGPT?

GPT-5.5 Instant becomes the new factory-default model that every ChatGPT user receives without additional configuration. Hallucination — the situation where a language model generates inaccurate or fabricated content presented as fact — is one of the most common criticisms of large language models, and OpenAI states it was a key focus of this update.

The second emphasis is personalization. OpenAI claims the model better tracks user context and tailors responses to cover user preferences and prior interactions.

What does the system card contain?

A system card is a document OpenAI releases alongside major models, containing evaluations, known limitations, and safety measures. A separate system card was published at a distinct URL for GPT-5.5 Instant, suggesting the model went through OpenAI’s standard safety assessment procedure before being set as the default.

Detailed benchmark results and comparisons with previous models were not publicly available beyond the brief RSS description at time of publication. Any specific performance claims should be verified directly against the system card.

Why does the Instant default change matter?

Changing the default matters because it changes what the vast majority of ChatGPT users experience without awareness. Users who do not manually select a model automatically switch to GPT-5.5 Instant — meaning changes in quality, behavior, and hallucination rates immediately affect millions of daily conversations.

For organizations building workflows on ChatGPT, the recommendation is to test existing prompts against the new default and verify they behave as expected.