GPT-5.5 Instant

Spec

Attribute	Value
Developer	OpenAI
Released	2026-05-07
Announced	2026-05-07
Context window	unknown
Pricing	unknown
License	unknown
Availability	ChatGPT; API (via chat-latest alias)
Positioning	ChatGPT's default fast-response model (cf. the "Instant" name)

Strengths

"smarter, clearer, and more personalized"

A claim of simultaneous improvement across three axes (intelligence, clarity, personalization).

Benchmarks

TBD — direct fetch blocked, needs supplementing

Compared To

• Predecessor: GPT-5 (page TBD)
• Announced the same day: new realtime voice models — voice reasoning/translate/transcribe

Significance

• A 0.5 minor version — an incremental release. However, the emphasis on the "Instant" line is a signal of latency differentiation.
• Announced after Claude Opus 4.7 and GPT-Rosalind (2026-04-16) — confirming a trend toward multi-layered model lineups

Sources

• OpenAI Spring announcements

June 25, 2026 Checkpoint Update

OpenAI released an updated GPT-5.5 Instant checkpoint on June 25, 2026 (source) (VentureBeat):

• Better at understanding the intent behind a question and adapting responses accordingly
• Improved handling of complex constraints (follow-up pushback handled properly, not repeated)
• Better shopping UX: location context, product recommendations, business info, images woven together
• Warmer conversational tone: fewer robotic lists, more intentionally designed responses
• Location context: surfaces truly relevant nearby options (restaurants, services)

API note: The chat-latest alias now points to this checkpoint — developers using chat-latest automatically receive the update without any code change. Rolled out to paid subscribers first, then free users.

This is an iterative checkpoint update — not a new model name. The first release was 2026-05-07.

Retired as the ChatGPT free default (2026-08-06)

GPT-5.6 Luna replaced GPT-5.5 Instant as the default model for ChatGPT Free and Go users, who also gained unlimited text chats (source).

GPT-5.5 Instant is also the baseline in the reliability figures OpenAI published with that change: responses containing at least one factual error on financial, medical and legal prompts were ~62% less common with Luna and ~68% less common with Sol than with GPT-5.5 Instant, on an OpenAI internal evaluation whose set is not published.

Nothing read states that the model was withdrawn from the API, or that the chat-latest alias moved. Only the ChatGPT free-tier default is stated to have changed.

Open Questions

• What does "Personalized" specifically mean? Integration with ChatGPT memory?
• Pricing tier?
• Is GPT-5.5 Instant still served on the API after losing the ChatGPT free default on 2026-08-06? No source read says either way.