On 18 October 2026, Google's US subscriptions page lists Google AI Plus at $4.99 a month, Google AI Pro at $19.99, and Google AI Ultra from $99.99, with a higher Ultra tier at $199.99. The Gemini API is billed per million tokens, and several of its published prices rise on set dates. For coding, the choice between the two is mostly about limits you cannot see versus meters you can.
This post uses Google's pages fetched on 18 October 2026. Google does not publish token allowances for the plans.
What the plans say about coding
| Plan | Published price | Coding wording |
|---|---|---|
| Free | $0 | Antigravity free tier: "basic weekly rate limits", with unlimited Tab completions and Command requests |
| Google AI Plus | $4.99 a month | "2x higher usage access than Free". No coding inclusion listed |
| Google AI Pro | $19.99 a month | Jules: "higher limits". Antigravity: "entry rate limits to agent model" |
| Google AI Ultra | From $99.99 a month | Jules: "highest limits". Antigravity "Ultra 5x" |
| Google AI Ultra, higher tier | $199.99 a month | Antigravity "Ultra 20x" |
Google also says the Gemini app on Pro gives 4x the free usage limits, and Ultra gives up to 20x the Pro limits. Those are chat limits. Every coding limit is described in relative terms ("entry", "higher", "highest"), never in tokens or requests. The page did not mention Gemini CLI or Code Assist, so this post says nothing about them.
What the API charges
Per million tokens, Standard tier, from ai.google.dev, fetched 18 October 2026:
| Model | Input | Output | Context cache | Notes |
|---|---|---|---|---|
| Gemini 4 Argon | $2.00 through 30 Nov 2026, then $4.00 | $10.00, then $20.00 | $0.10, then $0.20 | No free tier |
| Gemini 3.1 Pro Preview | $2.00 (up to 200K tokens), $4.00 above | $12.00, $18.00 above | $0.20 (up to 200K), plus storage | No free tier |
| Gemini 3.8 Flash | $0.75 through 31 Dec 2026, then $1.50 | $3.75, then $7.50 | $0.075, then $0.15, plus storage | Free tier available |
| Gemini 2.5 Flash-Lite | $0.10 | $0.40 | $0.01, plus storage | Free tier available |
Cached content on the Flash and Pro models also carries an hourly storage charge, listed on the page. The scenario below leaves storage out, so treat its totals as slightly low.
A scenario you can change
Inputs are mine, not Google data: per developer day, 3,000,000 cached input tokens, 350,000 fresh input tokens and 100,000 output tokens, over 22 working days.
prices = { # USD per MTok: fresh in, cached in, out
"Gemini 4 Argon, until 30 Nov": (2.00, 0.10, 10),
"Gemini 4 Argon, from 1 Dec": (4.00, 0.20, 20),
"Gemini 3.1 Pro Preview": (2.00, 0.20, 12),
"Gemini 3.8 Flash, to 31 Dec": (0.75, 0.075, 3.75),
"Gemini 3.8 Flash, from 1 Jan": (1.50, 0.15, 7.5),
}
cached, fresh, out, days = 3.0, 0.35, 0.1, 22
for m, (i, c, o) in prices.items():
day = fresh*i + cached*c + out*o
print(f"{m}: ${day:.2f} a day, ${day*days:.2f} a month")
That gives about $44.00 a month for Argon now and $88.00 after 1 December, $55.00 for 3.1 Pro Preview, and $18.97 then $37.95 for 3.8 Flash. Because the published price steps are dated, a budget built today should state which date it assumes.
How to choose
- A plan makes sense for an individual who wants a fixed bill and is happy with limits expressed as tiers.
- The API makes sense for CI, shared tools, and any team that wants per token control or needs a paid tier.
- Data terms differ. Google's API terms say paid services do not use your prompts to improve its products, while unpaid services use content "to provide, improve, and develop" Google products and may involve human review. See the free tiers post in this series.
Safeguard lets you bring your own model, so a Gemini key can sit behind your own agent while Safeguard's MCP tool calls stay unlimited.
Quick check
Why does the same Gemini workload cost different amounts in November and December?
Google lists dated price steps. Gemini 4 Argon moves from $2.00 to $4.00 per million input tokens on 1 December 2026, per the pricing page.
