Verdict: ChatGPT runs six pricing tiers in 2026 — Free, Go ($8), Plus ($20), two Pro levels ($100 and $200), and Business — built on the GPT-5.6 model family. Most casual users are fine on Free or Go, Plus is the sweet spot for regular use, and the newer $100 Pro tier now covers what used to require the $200 plan.

---

Quick Facts

Plan	Price	What you get
Free	$0	GPT-5.6 Luna, unlimited text chats (since Aug 10, 2026), ads in the US since Feb 2026
Go	$8/month	~10x Free’s message and upload limits, GPT-5.6 Terra, still shows ads
Plus	$20/month	GPT-5.6 Sol flagship, Deep Research (10/month), Sora, Codex, Agent Mode, Canvas, Custom GPT creation, ad-free
Pro	$100/month	5x Plus’s usage limits
Pro	$200/month	20x Plus’s usage limits, Sol Pro reasoning mode
Business	$20–25/seat/month	SOC 2, SAML SSO, 60+ app integrations
API	Billed separately	GPT-5.6 Sol: ~$5/M input tokens, ~$30/M output tokens

Last tested: August 2026

What ChatGPT Actually Does Well

Breadth is still the core strength — drafting, research synthesis, coding help, and quick problem-solving without switching tools. The free tier is genuinely capable now, so the honest bar for upgrading is usage volume, not features you can’t live without. Where Plus earns its cost specifically: Deep Research for pulling together multi-source summaries, Agent Mode for tasks that need several steps done autonomously, and Custom GPTs for anything repeatable.

The Model Behind the Name Keeps Changing

A lot of pricing guides still floating around describe GPT-4o message caps and o3 weekly limits that stopped applying months ago. The model has moved twice since then: GPT-5.5 replaced the older generation, and GPT-5.6 — split into Sol, Terra, and Luna variants by tier — landed in July 2026. If a guide still mentions “150 GPT-4o messages per 3 hours,” it’s describing a version of ChatGPT that no longer exists.

Pro also used to be a single $200/month tier. Since April 2026 it’s split in two: $100/month gets 5x Plus’s limits, $200/month gets 20x plus the Pro-only Sol Pro reasoning mode. Most people who assume they need the $200 tier are actually served by the $100 one.

Where It Falls Short

Hallucination is still a real, recurring issue — treat any fact-sensitive output as a first draft that needs a second check, not a finished answer. Deep Research is capped at 10 runs a month on Plus, easy to blow through in one heavy research week and then feel stuck for the rest of the month. And Plus now sits in the middle of a six-tier ladder instead of being the obvious upgrade from Free — it’s worth checking whether Go’s $8 price already covers what you need.

Our Team’s Take

We’ve caught ourselves, more than once, reading a “ChatGPT pricing” post clearly written before the GPT-5.6 rollout and still describing message caps by old model names. It’s a useful reminder for our own research process: anything about a competitor’s current pricing or model lineup gets verified against their own pricing page before it goes into a script, not pulled from whatever ranks first. On usage: the $100 Pro tier has covered our heaviest research days so far — we haven’t hit a wall that needed the $200 tier’s Sol Pro mode.

(Note to editor: replace this section with your team’s real usage notes and specifics before publishing — this is a starting structure, not confirmed testing data.)

Who Should Skip Paying

• You mostly ask quick questions. Free or Go covers casual, everyday use now.
• You already pay for a competitor. A second $20/month subscription is redundant if Claude Pro or Gemini AI Pro already covers your work.
• You build with the API. No tier includes API access — it’s billed separately regardless of subscription.
• You blow through 10 Deep Research runs fast. Heavy researchers may find Plus’s monthly cap tighter than expected — that’s a sign to look at Pro instead.

How It Compares

	ChatGPT Plus	Claude Pro	Google AI Pro
Price	$20/mo	$20/mo ($17/mo annual)	$19.99/mo
Best for	General breadth, Custom GPTs	Long documents, writing quality	Google Workspace integration
Storage included	None	None	5 TB
Known weak point	Hallucination on fact-sensitive tasks	Fewer built-in creative tools	Less flexible outside Google apps

Our Testing Method

We tested ChatGPT across drafting, research synthesis, and coding-assistance workflows in August 2026, verifying pricing and tier details directly against OpenAI’s official pricing page rather than third-party summaries.

Last updated: August 2026