ChatGPT Reasoning Effort: What Replaced Instant, Thinking, and Pro (2026)

ChatGPT no longer has an Instant, Thinking, and Pro mode picker. OpenAI retired that split on August 6, 2026, and since October 7, 2026 the model in Chat is GPT-6. On Plus, Pro, Business, and Enterprise, GPT-6 runs on GPT-6 Sol and handles both quick replies and longer reasoning, and a thinking level (Instant, Medium, and High on Plus, plus Extra High on Pro, Business, and Enterprise) controls how much reasoning it spends per answer. On Free and Go, GPT-6 runs on GPT-6 Luna and a Think button turns on deeper reasoning for a single message. Pro is the top thinking level on Pro, Business, and Enterprise plans, running GPT-6 Pro (powered by GPT-6 Astra), and GPT-5.6 Sol Pro remains an option there.

If you learned ChatGPT any time before August 2026, you learned a composer with a mode picker in it. You chose Instant for speed, Thinking for step-by-step work, Pro for the hardest problems. That mental model is now out of date, and most guides still teach it.

This page explains what actually replaced those modes, what the reasoning effort control does at each level, and how to get the behaviour you used to get from Thinking and Pro.

What Changed on August 6, 2026

OpenAI folded ChatGPT's separate fast model and reasoning model into a single GPT-5.6 Sol experience for paid users. Instead of routing your message to one of two models depending on which mode you picked, Sol handled both quick replies and extended reasoning inside one model.

The same announcement made three other changes that matter here:

• Text chats became unlimited on every plan, including Free. The message cap that shaped a lot of "use Instant to save your quota" advice no longer applies to text.
• Free and Go moved to GPT-5.6 Luna as the default model, replacing GPT-5.5.
• Accuracy improved measurably. OpenAI reported factual errors were 62% less common on GPT-5.6 Luna and 68% less common on GPT-5.6 Sol.

What Changed Again on October 7, 2026

OpenAI replaced GPT-5.6 with GPT-6 as the model in Chat, starting October 7, 2026 for Plus, Pro, Business, and Enterprise and October 8 for Free and Go. In the model menu it is labeled GPT-6. GPT-6 Sol powers it on paid plans at Instant, Medium, High, and Extra High, GPT-6 Luna powers it on Free and Go, and the Pro level uses GPT-6 Pro, powered by GPT-6 Astra, OpenAI's most capable model. GPT-5.6 Sol is still selectable on paid plans. GPT-6 also brought Intelligent UI, which lets answers include charts, diagrams, and interactive elements from Instant through Extra High. GPT-5.5 was retired from ChatGPT on October 14, 2026.

Old Mode to New Control: The Mapping

Thinking Levels (Plus, Pro, Business, and Enterprise)

Paid subscribers choose a thinking level in the ChatGPT composer, on web, mobile, and desktop, that sets how much thought GPT-6 puts into a response: Instant, Medium, and High on Plus, plus Extra High and Pro on Pro, Business, and Enterprise. It replaced the binary mode toggle with a graded control.

The practical rule is the same one that used to govern mode choice, just expressed differently:

• Stay on Instant for factual lookups, rewrites, summarising text you pasted, and ordinary conversation. Raising effort on these mostly buys you latency.
• Move to Medium for planning, writing that has to hold a structure, and code review. This is the range that used to be called Thinking.
• Go to High or Extra High when being wrong is expensive: multi-step maths, research synthesis across long sources, code that has to satisfy a specification, documents with hard compliance constraints.

Higher effort costs you time, and it is not unlimited: Medium, High, and Extra High draw on thinking usage limits, and the Pro level has its own allowance. On Business, GPT-6 Pro and GPT-5.6 Sol Pro share 15 Pro messages a month on Standard seats and 50 a week on Premium seats.

The Think Button (Free and Go)

The Think button is a per-message toggle, not a setting you turn on once. You tap it before sending a message that needs more care, and it applies to that message only.

The detail most guides get wrong: the Think button gives GPT-6 Luna more time to reason. It does not switch you to Sol. You are getting more deliberation from the same model, not access to GPT-6 Sol. If you need Sol itself, that requires a paid plan. Accounts the October rollout has not reached yet may still be on GPT-5.6 Luna.

OpenAI shipped it alongside unlimited text chats for Free and Go, subject to anti-abuse guardrails.

Plan Availability

Rollout was staged rather than instant, so a given account may have seen these controls appear over the days following August 6. GPT-6 likewise reached paid plans from October 7 and Free and Go from October 8, 2026, so an account the rollout has not reached may still show GPT-5.6.

The API Equivalent: reasoning.effort

If you are building on the API rather than using ChatGPT, the same idea is exposed as a parameter. The reasoning.effort parameter accepts seven settings:

none, minimal, low, medium, high, xhigh, max

This is the more precise version of what the consumer slider exposes, and it is the right place to look if you need reproducible behaviour across runs rather than a UI control someone might have nudged.

The GPT-6 and GPT-5.6 Families

Deliberately no price column here. OpenAI cut GPT-5.6 API prices twice in under a month, on July 30 and again on August 21, 2026, and published figures went stale between those dates. Read the current numbers from OpenAI's pricing page rather than from any article, including this one.

Context windows need the same care. OpenAI's model documentation lists 1.05M tokens of context and 128K of output for the GPT-6 and GPT-5.6 API models, but the ChatGPT app gives you far less: 27K tokens on Free, 54K on Go and Plus, and 128K on Pro for quick replies, rising to 256K on Go and Plus and 400K on Pro with reasoning.

If You Still See the Old Labels

A few things explain a composer that does not match this page:

• Staged rollout. The changes reached accounts over a period rather than all at once.
• Older documentation. A large amount of third-party writing, including tutorials published well after August 2026, still describes the three-mode picker. The screenshots are the giveaway.
• The model picker moved. Some users reported losing the quick in-composer way to switch models and having to go into settings instead. If you are hunting for a control you remember being in the composer, check settings first.

Legacy Naming, For Reference

These labels appear in older documentation and community posts. All of them predate the August 2026 change:

• Quick Response became Instant, which is now simply low reasoning effort.
• Think Deeper became Thinking, which is now raised reasoning effort.
• Pro mode is now the Pro thinking level, running GPT-6 Pro (powered by GPT-6 Astra), with GPT-5.6 Sol Pro still available on eligible plans.

How AI Toolbox fits alongside the new controls

The reasoning control is OpenAI's, and it works well. What ChatGPT still does not show you is any history of your own usage: how many messages you sent, how many of them used reasoning, and when. OpenAI documents its models and how they are developed in the OpenAI Help Center.

AI Toolbox (formerly ChatGPT Toolbox) keeps a local usage record: a daily count of messages sent, how many used reasoning versus a plain reply, and a dashboard to review it over time. It sits alongside the native controls rather than replacing them. The same unified install adds:

• Prompt library: save prompts and trigger them with the // shortcut in the composer.
• Folders, search, bookmarks, and pin: organise and find past conversations, including the long high-effort threads worth revisiting.
• Export, bulk archive, and bulk delete: keep history tidy after heavy research sessions.

One caveat worth stating plainly: the usage tracker was built against the Instant, Thinking, and Pro labels, so its per-mode breakdown is less reliable on accounts that have moved to the effort slider. The message and reasoning counts still record, the mode labels are the part that is catching up.

AI Toolbox is one unified Chromium install (Chrome, Edge, Brave, Opera, Arc). It does not run on Firefox.

Frequently Asked Questions

Does ChatGPT still have Instant, Thinking, and Pro modes?

No. OpenAI retired the three-mode picker on August 6, 2026, and since October 7, 2026 the Chat model is GPT-6. On paid plans, GPT-6 runs on GPT-6 Sol and handles both quick replies and extended reasoning, with a thinking level (Instant, Medium, High, and Extra High on Pro, Business, and Enterprise) controlling how much reasoning it spends. On Free and Go, GPT-6 runs on GPT-6 Luna and a Think button enables deeper reasoning per message. Pro is the top thinking level on Pro, Business, and Enterprise plans, running GPT-6 Pro (powered by GPT-6 Astra), with GPT-5.6 Sol Pro still available.

What replaced the Thinking mode?

The Medium, High, and Extra High thinking levels on paid plans, and the Think button on Free and Go. Choosing Medium or High before you send is the direct equivalent of picking Thinking. The Think button is the Free and Go equivalent, applied one message at a time.

Does the Think button give me GPT-6 Sol?

No. The Think button gives GPT-6 Luna more time to reason on that message. It does not switch you to Sol. Access to Sol requires a paid plan. This is the most common misunderstanding about the Free tier control.

Do reasoning modes still count against a message limit?

Text chats became unlimited on every ChatGPT plan, including Free, in the same August 6, 2026 update. Medium, High, and Extra High still draw on thinking usage limits, and GPT-6 Pro has its own allowance; on Business, GPT-6 Pro and GPT-5.6 Sol Pro share 15 Pro messages a month on Standard seats and 50 a week on Premium seats. Free and Go usage remains subject to anti-abuse guardrails, and workspace admins can set their own limits on Business and Enterprise.

Which model does each ChatGPT plan use now?

Every plan uses GPT-6 in Chat: Free and Go run it on GPT-6 Luna, and Plus, Pro, Business, and Enterprise run it on GPT-6 Sol, with GPT-6 Pro (powered by GPT-6 Astra) and GPT-5.6 Sol Pro available on Pro, Business, and Enterprise. Paid plans can still select GPT-5.6 Sol. GPT-6.1 Sol, GPT-6.1 Terra, GPT-5.6 Terra, and GPT-5.6 Luna are in ChatGPT Work and Codex only.

What is the API equivalent of the reasoning slider?

The reasoning.effort parameter, which accepts none, minimal, low, medium, high, xhigh, and max. Use it instead of the consumer slider when you need behaviour that is reproducible across runs.

Why does my ChatGPT still show the old mode picker?

The change rolled out over a period rather than all at once, so accounts saw it at different times. Note also that some users found the model picker moved out of the composer and into settings, so a control you remember may have relocated rather than disappeared.

Related Reading

• ChatGPT Models Explained: Complete Comparison Guide (2026), every current model with use cases.
• ChatGPT Limits: Messages, Tokens, and Rate Limits (2026), what still has a cap now that text chats do not.
• ChatGPT Keyboard Shortcuts: Complete Guide (2026).
• Save Reusable ChatGPT Prompts, the // shortcut and the saved prompt library.
• Automate ChatGPT Prompt Sequences, chaining multiple high-effort turns into one workflow.

Bottom Line

The question "Instant, Thinking, or Pro?" no longer has an answer, because the picker that asked it was retired on August 6, 2026. What replaced it is simpler: one model, GPT-6, and one control for how hard it thinks. Paid plans pick a thinking level. Free and Go tap Think. Text chats are unlimited either way, so the old habit of rationing your good messages is no longer necessary. If a guide still tells you to choose a mode in the composer, check its date.

References

Sources and documentation referenced in this article:

• Improving GPT-5.6 Sol in ChatGPT and expanding access to GPT-5.6 Luna for free users, OpenAI, August 6, 2026
• GPT-5.6: Frontier intelligence that scales with your ambition, OpenAI
• ChatGPT Release Notes, OpenAI Help Center
• Model Release Notes, OpenAI Help Center
• OpenAI Models, Platform documentation
• OpenAI brings unlimited ChatGPT text chats to free users, TechCrunch, August 6, 2026
• OpenAI drops ChatGPT text chat limits for free users, Help Net Security, August 7, 2026
• AI Toolbox (formerly ChatGPT Toolbox), Chrome Web Store listing

Retrieved August 2026, rechecked October 28, 2026. Last updated: October 28, 2026.