Gemini 4 Argon Is GA at $2/$10 Until 30 November: Our Verdict Moves to Conditional
- What happened
- Gemini 4 Argon went GA on the paid Gemini API on 13 October at $2/$10 per 1M tokens through 30 November 2026, then $4/$20.
- Why it matters
- We held Gemini 4 at pending because only Fairwind partners could use it. Paid API users can call it now, so it gets a real verdict: conditional.
- What to do
- Trial it on your own tasks before 30 November, and budget production at $4/$20 from 1 December.
Our verdict: Gemini 4 Argon moves from pending to conditional. We held it at pending because nobody outside Google's Fairwind cyber-defence partners could use it. Since 13 October anyone with a paid Gemini API key can, at a published price. It is conditional, not recommended, for three reasons: every benchmark is Google's own, we have not tested it, and the price doubles on 1 December 2026.
What happened
The Gemini API changelog entry for 13 October 2026 reads: "Gemini 4 Argon (gemini-4-argon) generally available (GA)". Google's launch post, by Koray Kavukcuoglu, says it "is now globally available today for developers and enterprises via the API in AI Studio and Gemini platform", and in the Gemini app for Google AI Ultra and Pro subscribers.
When Google announced Argon on 30 September, it gave no end date for the introductory price. Now it does. The launch post's footnote: "After the introductory period expires on November 30, 2026, the price of $4 per 1M input tokens and $20 per 1M output tokens will apply."
Google's pricing page, Standard tier, per 1M tokens:
| Through 30 Nov 2026 | From 1 Dec 2026 | |
|---|---|---|
| Input | $2.00 | $4.00 |
| Output (including thinking) | $10.00 | $20.00 |
| Context caching | $0.10 | $0.20 |
The free tier is "Not available", and caching is "Implicit caching only (explicit caching is not supported)". Google's model page lists gemini-4-argon as stable, with a 1,048,576-token context window shared between input and output and a default maximum output of 65,536 tokens.
Google says it is also "beginning expansion to hundreds of additional Fairwind Program partners", its programme for cybersecurity defenders.
Why it matters
The only reason Gemini 4 sat at pending was access. That reason is gone, so it gets a real verdict. It stops at conditional:
- Benchmarks are Google's. Its claims (first place on the Vals Index, 87% on ERQA, a 0.7% attack success rate on Gray Swan's indirect prompt injection test) all come from Google's own launch post. No independent numbers yet.
- We have not tested it. We will not imply experience we do not have.
- The price doubles on 1 December. At $4/$20 it costs twice the introductory rate, and with no free tier every thinking token is billed.
What changes for you
- Trial it now. The $2/$10 window runs through 30 November 2026. Run
gemini-4-argonon your own tasks before then. - Re-cost at $4/$20. Any budget built on the introductory rate is wrong from 1 December. Cached input moves from $0.10 to $0.20.
- Plan for implicit caching only. Explicit caching is not supported, so prompt layouts that depend on it will not carry over.
FAQ
Is Gemini 4 Argon free to try? No. Google's pricing page lists the free tier as "Not available", so every call, including thinking tokens, is billed at the paid rate.
When does the price change? The $2/$10 introductory rate runs through 30 November 2026. From 1 December 2026 it is $4 per 1M input tokens and $20 per 1M output tokens.
Why conditional and not recommended? Every benchmark so far is Google's own and we have not tested the model. We will revisit the verdict when independent results or our own runs are in.
Gemini 4 Argon
Gemini 4 Argon is generally available on the paid Gemini API since 13 October 2026, so the Fairwind-only reason for pending is gone. Conditional because every benchmark is Google's own, we have not tested it, and the introductory price doubles to $4/$20 on 1 December 2026.
What to do
- 1 Test gemini-4-argon on your own tasks while the introductory $2/$10 rate runs, through 30 November 2026.
- 2 Budget production at $4 input and $20 output per 1M tokens from 1 December 2026, and $0.20 for cached input.
- 3 Plan for implicit caching only: Google's pricing page says explicit caching is not supported, and there is no free tier.
Affected tools & models
Never need to catch up again
The weekly delta: only verdict changes and act-now items. No digest filler.