ChatGPT Models

As of Nov 6, 2026, ChatGPT's Chat tab runs GPT-6 (GPT-6 Sol on paid plans, a Luna model on Free and Go) with Instant, Medium, High and Extra High levels plus Pro, powered by GPT-6 Astra; ChatGPT Work and Codex offer GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Sol and GPT-6 Luna.

On Nov 6, 2026, ChatGPT answers with GPT-6 in the Chat tab: GPT-6 Sol powers Instant, Medium, High and Extra High, and Pro uses GPT-6 Pro, powered by GPT-6 Astra on paid plans, while Free and Go get a Luna model and a Think button. The longer-running ChatGPT Work and Codex tabs have their own list: GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Sol and GPT-6 Luna "are not available in regular ChatGPT conversations", and GPT-6 Astra is there too. Which models you see depends on the tab and on your plan.

Three people in different workplaces with laptops, with the words Introducing GPT-5.6 in ChatGPT OpenAI

Quick facts

Chat tab, paid plans	GPT-6 Sol for Instant to Extra High; GPT-6 Pro (GPT-6 Astra) for Pro, as of Nov 6, 2026
Chat tab, Free and Go	GPT-6 Luna per the Oct 7, 2026 release note and the pricing page; GPT-5.6 Luna per the help articles (OpenAI's pages disagree)
Work and Codex	GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Luna, plus GPT-6 Sol and the GPT-5.6 models
Picker levels	Instant, Medium, High, Extra High, Pro Standard, Pro Extended (since Jun 10, 2026)
Still selectable from GPT-5	GPT-5.6 Sol (paid), GPT-5.6 Sol Pro (Pro)
Codex default	`gpt-6.1-sol` in `config.toml`

Which models are in the Chat tab?

Since Oct 7, 2026, the Chat tab's default is GPT-6. On Plus, Pro, Business and Enterprise it is powered by GPT-6 Sol, and on Free and Go by a Luna model. In the model menu, "GPT-6 replaces Latest", and choosing it gives four effort levels backed by GPT-6 Sol and a Pro option backed by GPT-6 Pro, which runs on GPT-6 Astra (OpenAI Help Center). GPT-6 answers in the Chat tab can use Intelligent UI, interactive elements inside the reply, from Instant through Extra High, but the Pro option does not support Intelligent UI.

The Free and Go model is the one point where OpenAI's own pages disagree. The Oct 7, 2026 release note, OpenAI's announcement and the pricing table say Free and Go use GPT-6 Luna. The model help article and the Free Tier FAQ, both updated in the first days of November, say GPT-5.6 Luna. Either way it is a Luna model, the efficient end of the family (see GPT-6 Luna).

GPT-5.6 Sol stays as a separate choice in the menu for paid plans, and Pro subscribers also get GPT-5.6 Sol Pro (OpenAI Help Center). The pricing table still lists "Legacy models" for Plus and Pro and "GPT-5 Thinking Mini" on every plan, as of Nov 6, 2026. GPT-5.6 Terra and GPT-5.6 Luna are not selectable in ordinary chats. For everything that has left the menu, see Retired ChatGPT Models.

The model picker: Instant, Medium, High, Extra High and Pro

The picker's levels were renamed on Jun 10, 2026: Thinking Standard became Medium, Thinking Extended became High, Thinking Heavy became Extra High, and Thinking Light was dropped. What each one does, in OpenAI's words (OpenAI Help Center):

Level	What it is	Plans
Instant	"Fast responses for everyday questions"	All plans
Medium	"Standard thinking"	Plus, Pro, Business, Enterprise
High	"Extended thinking"	Plus, Pro, Business, Enterprise
Extra High	The highest thinking level with GPT-6 Sol	Pro, Business, Enterprise (not Plus)
Pro (Standard and Extended)	GPT-6 Pro on eligible plans, or GPT-5.6 Sol Pro	Pro, Business, Enterprise (not Plus)

Source: GPT-6 and other models in ChatGPT, OpenAI Help Center, read Nov 6, 2026.

On the web the picker sits in the message box (moved there on Apr 28, 2026); on iOS and Android it is at the top of the conversation. On the phone apps, paid users can long-press the send button to send one message with another model without changing the default. A Thinking slider sets the effort.

ChatGPT no longer moves your question up a level by itself. On Sep 14, 2026 OpenAI retired automatic switching from Instant to Thinking for Plus and Pro and removed the "Higher intelligence" setting on the web, though ChatGPT "can still switch automatically for safety purposes". On Enterprise and Edu, Instant no longer auto-switches either. If you want deeper reasoning, pick Medium or higher yourself.

The Think button on Free and Go

Free and Go have no effort levels. Instead they have a Think button for harder questions, introduced in August 2026 (on the web from Aug 14). "Think uses GPT-5.6 Luna, not GPT-5.6 Sol," says the help article, and Go "does not include GPT-5.6 Sol" (What is ChatGPT Go). Everyday text chats are unlimited on both plans, subject to abuse safeguards; files, images, voice and other tools keep separate limits. See Is ChatGPT Free and ChatGPT Go.

What happens when you hit a limit?

OpenAI publishes no message counts for the Chat models (ChatGPT Usage Limits has what is published). It does say what takes over. "A Pro subscription does not include unlimited use of the GPT-6 Pro model": on Pro $200, after the weekly GPT-6 Pro limit ChatGPT switches to GPT-5.6 Thinking at Medium, and on Pro $100, GPT-6 Pro and GPT-5.6 Sol Pro share one weekly allowance (OpenAI Help Center). On Plus, an Oct 29, 2026 note refers to "the model ChatGPT can switch to after you reach GPT-6 message limits" without naming it (Model Release Notes). Earlier in 2026 OpenAI named small fallback models in the release notes, most recently GPT-5.5 Instant Mini on Jul 6, 2026; no later page read on Nov 6, 2026 says whether that still applies.

Which models are in ChatGPT Work and Codex?

ChatGPT Work and Codex use a separate model list, unchanged by the Chat tab's switch to GPT-6. OpenAI's Codex docs recommend four (learn.chatgpt.com):

Model	OpenAI's description	Who gets it
GPT-6 Astra	"Our most capable model for complex work across code, apps, and research"	Full allowance on Pro $100, Pro $200 and Business Premium; limited on Plus and Business Standard
GPT-6.1 Sol	"Near-Astra performance for complex work at a lower cost than Astra"	Plus, Pro, Business, Enterprise, Edu (off by default on Enterprise and Edu)
GPT-6.1 Terra	"A fast, efficient model for clear writing, visual understanding, and computer use"	Plus, Pro, Business, Enterprise, Edu
GPT-6 Luna	"Our most efficient model for focused, high-volume tasks"	Every plan; Free and Go in the desktop app

Sources: Codex models and Codex pricing, OpenAI, and ChatGPT Work and Codex, OpenAI Help Center, all read Nov 6, 2026.

GPT-6 Sol is still listed beside them, and the GPT-5.6 Sol, Terra and Luna models remain in Work for paid plans (OpenAI Help Center). GPT-6.1 Sol has a default context window of 500,000 tokens there. Ultrafast speed is limited to Pro $500 and eligible Enterprise and Edu. Codex's own effort names run Light, Medium, High, Extra High, Max and Ultra; see Codex Models.

How much can each model read at once?

The pricing page gives these context windows per plan, as of Nov 6, 2026; the part left for your own input is smaller than the total.

Plan	Instant	Thinking levels
Free	27K tokens	"Varies"
Go	54K tokens	256K tokens
Plus	54K tokens	256K tokens
Pro	128K tokens	400K tokens

Source: ChatGPT pricing, read Nov 6, 2026.

The same models in the OpenAI API accept far more: 1,050,000 tokens for every GPT-6 flagship.

Which model should you pick?

Everyday questions, quick drafts, lookups: Instant. It is the only level Free and Go plans have, and the fastest.
Math, analysis, careful writing, code you will paste somewhere: Medium or High. These are the levels a Plus plan includes.
The hardest problems in Chat: Extra High or Pro, on Pro, Business or Enterprise. Pro is GPT-6 Astra, OpenAI's most capable model, but it is rationed weekly and gives up Intelligent UI.
Long tasks that produce files, research reports or code changes: move to the Work or Codex tab. GPT-6.1 Sol is the everyday choice there; OpenAI pitches GPT-6.1 Terra for writing, images and computer use at lower cost, and GPT-6 Luna for high-volume, simple jobs.
A ranking against Claude, Gemini and others: see OpenAI Benchmarks. On the dated leaderboards read Nov 6, 2026, no OpenAI model is first overall.

What testers found

4:11 In a Sep 13, 2026 video comparing GPT-6 Astra in ChatGPT with Claude Fable 5.1, Lisa Zachmann asked both a simple question about the new iPhone and got the same correct answer from each. Her advice was to save the top models for hard tasks, since a lookup like that is little more than a web search: "do we use Albert Einstein to mop the floor?"

Frequently asked questions

What model does ChatGPT use right now? As of Nov 6, 2026, ChatGPT's Chat tab uses GPT-6: GPT-6 Sol on Plus, Pro, Business and Enterprise, with GPT-6 Astra behind the Pro option, and a Luna model on Free and Go. OpenAI's pages disagree on whether the Free model is GPT-6 Luna or GPT-5.6 Luna.

Is GPT-6 Astra in ChatGPT? Yes, in two places: the Pro option in the Chat tab (on Pro $100, Pro $200, Business and Enterprise) and the model list in ChatGPT Work and Codex, where Plus and Business Standard get limited use. It is not the default anywhere in Chat.

Why can't I find GPT-6.1 Sol in my chat? GPT-6.1 Sol and GPT-6.1 Terra are only in ChatGPT Work and Codex, not the Chat tab. Switch tabs in the desktop app, or open Codex.

Does ChatGPT pick the model for me? Not any more on Plus and Pro: automatic switching from Instant to Thinking was retired on Sep 14, 2026. You pick the level in the picker; ChatGPT may still switch for safety reasons.

Can I still use GPT-4o in ChatGPT? No. GPT-4o was retired from ChatGPT on Feb 13, 2026 and from custom GPTs in business workspaces after Apr 3, 2026. It remains in the API until Jan 14, 2027.

What is Extra High in ChatGPT? Extra High is the highest thinking level of GPT-6 Sol in the Chat tab, the old Thinking Heavy. Pro, Business and Enterprise include it; Plus does not.