Skip to content
Promtime

Guide

OpenAI models and products

Which OpenAI models and products are current

Updated 21 October 2026288 stories

In short

OpenAI’s current frontier lineup is GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna. On October 21 all three began rolling out in private preview to select enterprise customers on Google Cloud’s Model Garden. General availability there is planned in the coming weeks, and Microsoft remains OpenAI’s primary cloud partner. In ChatGPT, Pro users get GPT-6 Astra Pro, which can split one task among parallel agents in the Chat tab. Voice in ChatGPT, Codex and dots runs on GPT-Live-1.1, with API access “coming soon”. GPT-4o retires in January 2027, and OpenAI recommends GPT-6 Luna as its replacement.

Recent changes developers will notice: GPT-6.1 Sol now holds 500,000 tokens of context by default in Codex and ChatGPT Work, up from 272k. Logged-in Codex and Work users keep their prompt cache forever, and API developers get a 12-hour cache. OpenAI also shipped an official CLI for its REST API. Its output is readable text by default, so scripts that parse responses need --format json or jsonl. Image generation in the CLI defaults to gpt-image-2.5-sunburst, and images create-variation is retired. On the revenue side, ChatGPT Ads hit a 1 billion dollar annualized run rate in under 200 days. Ads run in a separate service, so a slow ad server can’t delay ChatGPT’s reply.

At a glance

Model or productWhat it isStatusAs ofSource
OpenAI models on Google Cloud (GPT-6 Astra, GPT-6.1 Sol, GPT-6 Luna)OpenAI frontier models and Codex agentic workflows in Google Cloud's Model Garden, bought through Google Cloud MarketplaceRolling out in private preview to select enterprise customers; general availability planned in the coming weeks21 October 2026Read
GPT-6.1 SolOpenAI model; default context 500,000 tokens in Codex and ChatGPT Work (previously 272k)Available in Codex and ChatGPT Work with 500k default context12 October 2026Read
GPT-6.1 Sol at 8x speedFaster GPT-6.1 Sol option at 60 dollars per million output tokens8 October 2026Read
GPT-6 Astra ProChatGPT model that splits one task among parallel agents, with faster responses and better use of plugins and toolsAvailable to Pro users in the Chat tab only16 October 2026Read
GPT-6 LunaOpenAI's recommended replacement for GPT-4.x workloads; lower per-token price than the GPT-4.x models the migration coversRecommended migration target14 October 2026Read
GPT-4oOlder GPT-4.x modelHeading for retirement in January 202714 October 2026Read
GPT-Live-1.1Voice model powering Voice in ChatGPT, Codex and dots; new voices in French, German, Hindi, Mexican Spanish, Korean and JapaneseRolling out to all users; API access "coming soon"15 October 2026Read
OpenAI CLIOfficial command-line tool for the OpenAI REST API; brew install openai/tools/openai, or Go 1.26 or later; openai images generate defaults to gpt-image-2.5-sunburstReleased; images create-variation retired in favor of images edit with a GPT Image model21 October 2026Read
Prompt cachingStores the already processed part of a promptLogged-in Codex and Work users: cache kept forever; API: 12-hour cache21 October 2026Read
ChatGPT Pro 500New Pro plan with 8x faster Codex29 September 2026Read
ChatGPT Pro (200 dollars)Pro planUsage cut in half; 500 dollar tier added30 September 2026Read
ChatGPT AdsAds in ChatGPT, run by a custom post-trained LLM orchestrator in a separate service, with an auction tool that picks the winner and sets the price1 billion dollar annualized revenue run rate less than 200 days after launch; more than 100,000 advertisers21 October 2026Read
Show all (25)
Model or productWhat it isStatusAs ofSource
Locked ChatsProtected space for ChatGPT chats, opened with a PIN on web and desktop, or with a PIN, Face ID, Touch ID or device unlockRolling out to Free, Go, Plus, Pro, Business, Enterprise and Edu plans20 October 2026Read
Rosalind / GPT-RosalindOpenAI's life sciences models and tools; GPT-Rosalind reasons in biology and chemistryOrchestra Bio's team and technology join Rosalind after the acquisition20 October 2026Read
DotsOpenAI agentsRun free until they open a Codex task1 October 2026Read
Codex in the cloudCodex keeps coding in the cloud after your laptop closes29 September 2026Read
Codex Security CloudReviews commits with the laptop closed30 September 2026Read
Codex third-party modelsKimi K3 and GLM-5.3 Flash in Codex via Baseten30 September 2026Read
Codex limit resets"Allow Codex to use resets" setting under Usage & billing; Codex can spend a saved reset when asked mid-conversation and the current limit is at 10% or lessResets are never used automatically10 October 2026Read
ChatGPT and Codex desktop appsDesktop apps with a new storage scheme for app stateNew storage scheme should reduce data kept on disk by 60%15 October 2026Read
ChatGPT Plus and Pro usage in third-party appsChatGPT Plus and Pro plans can pay for other apps' AI30 September 2026Read
MCP ExtensionsPut plugins in the ChatGPT sidebar30 September 2026Read
ChatGPT SpacePuts people and agents in one document29 September 2026Read
Decisions APIOpenAI API that picks an answer in 150 milliseconds29 September 2026Read
textGrainText watermark: a secret key nudges each token pick, and a detector with the same key runs a statistical testMarks ChatGPT and Codex text by default in the EU10 October 2026Read

Milestones

  1. OpenAI puts 8x faster Codex behind a new Pro 500 plan.
  2. Codex keeps coding in the cloud after your laptop closes.
  3. ChatGPT Space puts people and agents in one document.
  4. OpenAI cuts usage on the 200 dollar Pro plan in half and adds a 500 dollar tier.
  5. OpenAI’s dots work 24/7 on their own cloud computer.
  6. ChatGPT Plus and Pro usage opens to third-party apps.
  7. Codex opens up to Kimi K3 and GLM-5.3 Flash via Baseten.
  8. MCP Extensions put plugins in the ChatGPT sidebar.
  9. OpenAI speeds up GPT-6 models by about 50% in ChatGPT.
  10. GPT-6.1 Sol gets 8x speed at 60 dollars per million output tokens.
  11. textGrain marks ChatGPT and Codex text by default in the EU.
  12. GPT-6.1 Sol’s default context in Codex and ChatGPT Work rises from 272k to 500,000 tokens.
  13. OpenAI names GPT-6 Luna the replacement for GPT-4o, which retires in January 2027.
  14. GPT-Live-1.1 starts powering Voice in ChatGPT, Codex and dots for all users.
  15. GPT-6 Astra Pro in ChatGPT splits one task among parallel agents for Pro users.
  16. Locked Chats roll out across ChatGPT plans.
  17. GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna enter private preview on Google Cloud.
  18. OpenAI ships an official CLI for its REST API.
  19. Codex and Work keep their cache forever, and the API gets a 12-hour cache.
  20. ChatGPT Ads reach a 1 billion dollar annualized run rate in under 200 days.

FAQ

What are OpenAI’s current GPT-6 models?

The models OpenAI is bringing to Google Cloud as its frontier models are GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna. They are rolling out in private preview to select enterprise customers (source).

In ChatGPT, Pro users also get GPT-6 Astra Pro, which runs parallel agents in the Chat tab (source).

When is GPT-4o being retired and what replaces it?

GPT-4o heads for retirement in January 2027, and OpenAI recommends GPT-6 Luna as the replacement. The migration guide says to start with reasoning effort set to none and to send only a limited share of production traffic at first. Luna has a lower per-token price than the GPT-4.x models the migration covers, but token usage and reasoning settings can affect the cost of each task (source).

What is GPT-6.1 Sol’s context window in Codex?

GPT-6.1 Sol’s default context in Codex and ChatGPT Work is 500,000 tokens, up from 272k. OpenAI plans to raise it further but hasn’t said when or to what number (source).

Can I use OpenAI models on Google Cloud?

Yes, in private preview for select enterprise customers. GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna are rolling out in Model Garden, and purchases go through Google Cloud Marketplace on a single Google Cloud invoice. General availability is planned in the coming weeks, and Microsoft remains OpenAI’s primary cloud partner (source).

How long does OpenAI keep the prompt cache?

For logged-in Codex and Work users, the cache is kept forever and they’re guaranteed to hit it. API developers get a 12-hour cache. This builds on the September 22 caching update for GPT-6, which cut cached input token costs by up to 90% (source).

Latest