GPT-5.5 changelog

GPT-5.5 retired from Codex and ChatGPT

At noon PDT (19:00 UTC) on October 14, 2026, GPT-5.5 retired from Codex for ChatGPT sign-in. Replacements are GPT-6.1 Sol on paid plans and GPT-6 Luna on Free and Go. API-key sessions are not affected.

Who is affected:

Codex and ChatGPT users and scheduled tasks that selected GPT-5.5.

Action needed:

Replace gpt-5.5 with gpt-6.1-sol in defaults, custom agents, scheduled tasks and scripts.

GPT-5.5 to retire from ChatGPT, ChatGPT Work and Codex on October 14

OpenAI announced GPT-5.5 would retire from ChatGPT, ChatGPT Work and Codex on all plans on October 14, 2026. The API is not affected.

Who is affected:

ChatGPT and Codex users and admins who select GPT-5.5.

Action needed:

Update workspace defaults, saved settings, custom agents and scripts that select gpt-5.5.

Codex users report GPT-5.5 reasoning tokens clustering at 516, linked to weaker answers

A GitHub issue analyzing Codex token metadata from February 1 to June 27, 2026 found gpt-5.5 responses landing disproportionately at exactly 516 reasoning tokens, consistent with a reasoning-budget threshold. The author says it does not prove truncation. The thread drew 372 points and 152 comments.

Who is affected:

Codex users running gpt-5.5 on complex tasks.

What to do:

No action needed. GPT-5.5 retired from Codex with ChatGPT sign-in on October 14, 2026.

OpenAI models available in Amazon Bedrock

GPT-5.4 and GPT-5.5 became available in Amazon Bedrock through an OpenAI-compatible Responses API endpoint, with models and features varying by AWS Region.

GPT-5.5 becomes the default orchestration model in Perplexity Computer

Perplexity began rolling out GPT-5.5 as the default orchestration model in Computer for Pro and Max subscribers. GPT Image 2 became the default for image generation and editing in Computer.

GPT-5.5 and GPT-5.5 Pro released in the API

GPT-5.5 shipped to the Chat Completions and Responses APIs, with GPT-5.5 Pro on the Responses API. Reasoning effort defaults to medium, and caching for GPT-5.5 only works with extended prompt caching.

Who is affected:

Developers choosing a frontier OpenAI model on the API.

What to do:

Test GPT-5.5 against your current model. Note that in-memory prompt caching is not supported.

GPT-5.5 available in Codex and recommended for most tasks

GPT-5.5 became available in Codex as the recommended choice for most Codex tasks once it appears in the model picker.

Who is affected:

Codex users.

What to do:

Select gpt-5.5 in Codex.