Skip to main content
SkillTycoonSign in

Lesson 1 of 9 · Foundations - ChatGPT power user

GPT-5.5 modes: Instant, Thinking, Pro

From Create with OpenAI: A beginner-to-advanced tour of OpenAI's stack: GPT-5.5 modes, Memory + Projects, Canvas + Tasks, GPT Image 2, the Realtime voice trio, Code Interpreter, Codex on CLI, Cloud and Mobile, and AgentKit — plus the limits, watermarking and refusal layer every shipper must own. Three modules, nine lessons.

Module 1 - Lesson 1

GPT-5.5 modes: Instant, Thinking, Pro

ChatGPT now ships with GPT-5.5 Instant as its default model and exposes two heavier siblings - Thinking and Pro - for harder asks. Picking the right mode is the highest-leverage habit a serious ChatGPT user can build.
  • Level: Foundations
  • Time: ~15 min
  • Skill: Mode selection

ChatGPT's default model is now GPT-5.5 Instant. OpenAI's published numbers claim a meaningful drop in hallucinations on high-stakes prompts in medicine, law, and finance compared to the prior default - somewhere around half - while keeping the same fast response latency.

The bigger insight isn't the speed bump. It's that ChatGPT now exposes three distinct modes - Instant, Thinking, Pro - and most users still take whatever the picker last left them on. Choosing deliberately is the single highest-leverage habit of a serious ChatGPT user.

The default model matters

If you haven't touched the model picker in a while, ChatGPT may have silently upgraded you. That's a real improvement - but it also means prompt habits you built on the older default can produce subtly different output shapes today. Audit your saved prompts when output style drifts.

The three modes, side by side

Tap each card to see when to use it.
  • Instant - the default
  • Thinking - for hard problems
  • Pro - heavy research

Match each task to the strongest mode

Drag, then press Check.
Rewrite this 4-line email to sound warmer
Plan a 3-week migration from REST to gRPC across 12 services
Synthesise 14 medical-journal PDFs into a clinician briefing
Translate a paragraph from Hindi to English

The decision rule

  • Reach for Instant
  • Reach for Thinking
  • Reach for Pro
  • When in doubt

Tuesday, 9:42 AM

You're staring at a one-line message from your manager: 'Can you prep a 5-min stand-up update on the Q3 onboarding redesign by 10 AM?' You have 18 minutes. ChatGPT is open in another tab. Which mode do you reach for?

The clock is real. The stakes are mild. Choose.

Pick the mode

Best fit?

Quick check

Two questions before we move on.
Which model is currently ChatGPT's default?
Which of these are valid reasons to reach for Thinking over Instant?

Before the next lesson

Open ChatGPT and look at the model picker. Confirm it currently reads GPT-5.5. Toggle between Instant, Thinking, and Pro at least once - just to know where each one lives. In the next lesson we'll wire Memory and Projects on top.

Topics: openai, chatgpt, gpt-5-5, codex, agentkit, gpt-image-2, realtime-voice, canvas, tasks, create-with-openai