Unofficial guide. Not affiliated with Google.
Last verified: 21 October 2026
What Gemini 4 Argon is
Gemini 4 Argon (model ID: gemini-4-argon) was announced on 30 September 2026 to Fairwind cybersecurity partners only, and became generally available on 13 October 2026.
It has a context window of 1,048,576 tokens, three thinking levels — low, medium and high — and implicit caching only.
Access: who can use it, and where
- Gemini app and AI Mode — available to Google AI Pro ($19.99/mo) and Ultra subscribers.
- Gemini API — paid only. There is no free tier.
- A Pro subscription does not include the API — Google AI Pro does not include Gemini API or AI Studio access. For API access you need Ultra, or a paid API key.
- Antigravity — Ultra-only.
Pricing, and the 1 December price rise
These prices are confirmed from Google’s pricing page, via multiple sources. All prices are per 1M tokens:
| Tier | Through 30 November 2026 | From 1 December 2026 |
|---|---|---|
| Standard — input / output | $2 / $10 | $4 / $20 |
| Cached input | $0.10 | $0.20 |
| Batch / Flex — input / output | $1 / $5 | $2 / $10 |
| Priority — input / output | $3.60 / $18 | $7.20 / $36 |
In other words, standard pricing doubles on 1 December 2026: from $2/$10 per 1M input/output tokens through 30 November, to $4/$20 from 1 December.
One point that matters for cost: the output price includes thinking tokens, so the reasoning tokens the model generates are billed as output. More thinking means a higher bill, even where the visible answer is short.
Benchmarks
On the Artificial Analysis Intelligence Index, Gemini 4 Argon scores 53 — tied with GPT-6 Astra.
Google has also published a vendor-reported table of 19 benchmarks. In that table, Argon is first or tied on 14 of the 19, leading on DeepSWE v1.1 at 77.9%, the Vals Index at 68.9% and Harvey Legal at 19.6%. It trails on FrontierSWE v2 at 55.0% (GPT-6 Astra: 65.5%) and on Terminal-Bench 4.0 at 57.4% (Opus: 66.4%). Those figures are Google’s own vendor-reported numbers.
FAQ
When did Gemini 4 Argon become generally available?
13 October 2026. It was announced earlier, on 30 September 2026, to Fairwind cybersecurity partners only.
Is there a free tier for Gemini 4 Argon on the API?
No. The Gemini API is paid only — there is no free tier.
Does a Google AI Pro subscription include API or AI Studio access?
No. Pro covers the Gemini app and AI Mode, but a Pro subscription does not include Gemini API or AI Studio access. You need Ultra, or a paid API key.
Who can use Antigravity?
Ultra subscribers only.
What happens to pricing on 1 December 2026?
Standard pricing doubles from $2/$10 to $4/$20 per 1M input/output tokens. Cached input rises from $0.10 to $0.20, Batch/Flex from $1/$5 to $2/$10, and Priority from $3.60/$18 to $7.20/$36.
Are thinking tokens billed?
Yes. The output price includes thinking tokens, so reasoning tokens are billed as output.
What is Argon’s context window?
1,048,576 tokens, with thinking levels of low, medium and high, and implicit caching only.
Argon not showing up?
If you are paying for Pro or Ultra but Argon is missing where you expect it, the problem is usually one of the three access gates. See the companion guide: Gemini 4 Not Showing Up? The 3 Access Gates Explained (App, API, Antigravity).
Last verified: 21 October 2026
Unofficial guide. Not affiliated with Google.
Tools and comparisons
Estimate your API bill with the Gemini API Cost Calculator, thinking tokens included. See how Argon compares with Meta’s model in Muse Wonder vs Gemini 4 Argon: Price, Benchmarks and Access Compared.
Full pricing table
Every Argon rate – Standard, Batch/Flex, Priority and cached input, before and after the 1 December 2026 doubling – is tabulated on the Gemini 4 Argon pricing page.
Is any of it free?
Argon itself is not free on any route – the short version is in Is Gemini 4 Free? No – What Google Actually Charges for Argon. For the genuinely free options around all three October launches, see Free Alternatives to Muse Wonder, Gemini 4 Argon and FLUX 3 Max.