The Gemini app — the free tier runs on the fast Flash models, with the Pro model and higher limits reserved for paid plans.
Photo: Wikimedia Commons (public domain).
In 60 seconds: Google now sells five tiers of Gemini — Free, AI Plus ($4.99), AI Pro ($19.99), and two Ultra plans ($99.99 and $199.99) — and the real differences aren't the prices, they're the message limits, the models you unlock, and the context window. Here's every number that matters, verified against Google's own plan pages.
- Gemini's free tier is genuinely useful — fast Flash models, image generation, Deep Research, voice, and Canvas — but its strongest models and the 1-million-token context window are now reserved for paid plans.
- Google AI Pro ($19.99/month) is the sweet spot for most people: the Pro model, 1M-token context, 1,000 video credits, 5 TB of storage, and roughly 4x the free tier's usage.
- Google AI Plus ($4.99/month) is the new entry tier: about 2x free usage, 400 GB of storage, and video generation for casual users.
- The two Ultra plans ($99.99 and $199.99) buy 5x and 20x the Pro usage respectively, plus YouTube Premium, Deep Think reasoning, and up to 30 TB of storage.
- Google doesn't publish exact message counts — limits are compute-based, refresh roughly every 5 hours, and shift with server load. If an "old Gemini Advanced" page confuses you, that's now Google AI Pro.
Everyone asks the same question the week they start using Gemini: when do you actually need to pay? The answer is less obvious than it should be, because Google renamed its plans in 2026 (goodbye, "Gemini Advanced" — hello, "Google AI Plus") and never publishes exact message caps. Here's the plain-English version, with every tier laid out side by side.
What's free — and what quietly isn't
Let's start with the good news: Gemini's free tier is one of the strongest free AI plans out there. You get the fast Flash-family models with automatic model selection, real-time voice conversations, image generation and editing, Deep Research reports, the Canvas side workspace, custom Gems, and 15 GB of standard Google storage. For casual questions, short documents, and everyday image work, most people never hit a wall.
Here's the catch, and it's the part that trips people up: Google has been moving its strongest models into paid tiers through 2026. The free tier now runs on the faster, cheaper models — if you want Gemini 3.1 Pro, Gemini 4 Argon, or Deep Think reasoning, you're paying. And while Google publishes the context window for paid plans, free-tier users get a smaller window, which shows up when you try to feed Gemini a very long document or a whole codebase.
The numbers: every tier, side by side
These figures are compiled from Google's own plan pages (gemini.google/subscriptions and Google's AI plan list) and independent testing, current as of October 2026. One caveat worth repeating: Google doesn't publish exact message counts — usage limits are compute-based, reset roughly every 5 hours, and tighten when servers are busy. Treat message figures as tested estimates, not guarantees.
| Plan | Price | Models | Context window | Usage | Image & video | Storage & extras |
|---|---|---|---|---|---|---|
| Free | $0 | Flash-Lite / Flash (auto-selected) | ~32k tokens | Standard; strictest rate limits | Image generation included; limited video | 15 GB; Deep Research (limited), Live, Canvas, Gems |
| Google AI Plus | $4.99/mo | More access than Free | ~128k tokens | ~2x Free usage | Higher image & video limits; 200 Flow credits | 400 GB; scheduled actions, Daily Brief, family sharing |
| Google AI Pro | $19.99/mo | Gemini 3.1 Pro / 4 Argon, Deep Think access | 1M tokens | ~4x Free usage; priority during peaks | Imagen 4, Veo 3.1 Lite trial; 1,000 Flow credits | 5 TB; $10/mo Cloud credits; Gemini in Gmail/Docs; Jules & Antigravity raised limits |
| Ultra 5x | $99.99/mo | Pro models + priority | 1M tokens | 5x Pro usage | Veo 3.1; 10,000 Flow credits | 20 TB+; YouTube Premium; Project Mariner automation |
| Ultra 20x | $199.99/mo | Highest access, agent mode | 1M tokens | 20x Pro usage; minimal limits | Veo 3.1; 25,000 Flow credits | 30 TB; priority access; Project Genie; higher Cloud credits |
A couple of things people get wrong: the Gemini API is billed separately — per token, pay as you go — and isn't bundled with any app plan (though Pro throws in $10 of Cloud credits a month). And "5x" in "Ultra 5x" means five times the Pro usage ceiling, not five times the price. Love it or hate it, Google is at least consistent about the math.
Why it matters
Here's the thing: for most Americans, this decision is really between Free and Pro. The Plus tier is new in 2026 and honestly fills a small niche — people who use Gemini daily and want scheduled actions and a bit more headroom, but don't need the Pro model. Ultra only makes sense for a very specific crowd: heavy video creators burning through Flow credits, researchers running Deep Think all day, or teams automating workflows with Project Mariner.
The number that decides it for most people is the context window. Thirty-two thousand tokens on Free sounds like plenty until you try to drop a 200-page contract or a semester's worth of lecture notes into the chat box. A 1-million-token window on Pro handles entire codebases and long PDFs without breaking a sweat. If your use case is "big documents," the $19.99 pays for itself the first week.
What this means for you
Stay on Free if you ask questions, summarize articles, generate images, and use voice chat a few times a day. It's genuinely free and genuinely good.
Step up to Plus ($4.99) if you hit free-tier rate limits regularly and want scheduled actions or more video generation without paying $20.
Step up to Pro ($19.99) if you work with long documents or code, want the strongest Gemini models, or use Gemini inside Gmail and Docs all day. The 5 TB of storage alone is worth nearly $10 a month as a Google One plan, which makes the AI premium feel a lot smaller.
Step up to Ultra if you generate video professionally or run research agents for hours — and you can expense it. For everyone else, it's overkill.
Our take: Pro is the only tier most people should seriously consider paying for. Free is excellent, Plus is a half-step, and Ultra is priced for businesses that bill clients for the output.
What happens next
Google's AI lineup has moved fast in 2026 — four plan changes this year alone, per industry trackers — and the trend is unmistakable: the free tier keeps the speed and the basics, while the frontier features keep migrating upward. Gemini Omni (video generation baked into the app) is already rolling out to paid tiers, and a higher-end Gemini Omni Pro has been teased without a date.
The smart move: pick your tier for the model and context window you need today, not the features Google might ship tomorrow. And if Google renames everything again — which, given this year's track record, it will — check the plan page before you renew, not after.
Frequently asked questions
Is Google Gemini really free?
Yes — the free tier costs nothing with a Google account and includes the fast Flash models, image generation, Deep Research, voice conversations, Canvas, and custom Gems. The catch is that Google's strongest models (Gemini 3.1 Pro, Gemini 4 Argon, Deep Think) and the 1-million-token context window are reserved for paid plans, and free-tier users hit rate limits sooner.
What happened to Gemini Advanced?
It was renamed. "Gemini Advanced" is now Google AI Pro ($19.99/month) and the top tier became Google AI Ultra. If an older guide tells you to buy Gemini Advanced, it's describing today's Google AI Pro.
Do I need a paid plan to use the Gemini Pro model?
Yes. Through 2026, Google moved its pro-tier models out of the free tier. Free accounts run on the faster, cheaper Flash-family models; Gemini 3.1 Pro and above require Google AI Pro or Ultra.
Is the Gemini API free?
Not for real use. The API is billed separately per token (pay as you go), apart from the app subscription plans. Google AI Pro includes $10 a month in Google Cloud credits, which can cover light API experimentation.
Does Gemini train on my chats?
Google may use your conversations to improve its models while the "Keep Activity" setting is on; turning it off stops your chats from being saved to your activity and used for training.
Related reading: our explainers on ChatGPT's token limits, OpenAI's API rate limits, and Claude's Free vs Pro vs Max tiers.
Sources: gemini.google/subscriptions and Google's AI plan list (via Krater, updated Oct 2026); ToolScout Gemini review (Oct 2026); AI Toolbox Gemini vs ChatGPT comparison (Oct 2026); saascrmreview Gemini pricing (Sept 2026).