Google: what changed, and what it means for you
Gemini models, the Gemini API and Google AI plans. What changed, what it costs you, and what to do.
Google unveiled Gemini 4 Argon on September 30—its first Gemini 4 model—with more top benchmark scores than GPT-6 Astra or Claude Opus 5.5 on Google’s own tables, at an introductory $2 / $10 per million tokens. You cannot use it yet: cyber defenders go first, then paid API customers and Google AI Ultra.
+ 2 good for readers − 1 costs or risks
Good news and bad news, dated
Gemini 4 Argon: the most top scores, at an introductory $2 / $10
By VentureBeat’s count of Google’s tables, Argon leads or ties on 13 of 18 disclosed benchmarks against GPT-6 Astra and Opus 5.5, with responses up to 1M tokens long. These are Google’s own numbers.
For you: Developers: a frontier-class model at Sonnet-level price—once you can get it.
You can’t use it yet
First access goes to cyber defenders in Google’s Fairwind Program and to U.S. government pre-release testing. Paid API customers and Google AI Ultra subscribers come next, with no date; Argon is not on Google’s API pricing page.
For you: Don’t plan a migration around it this month.
The introductory price will double
After an unannounced introductory period, Argon moves to $4 input / $20 output per million tokens—the same list price as Claude Opus 5.5.
For you: Budget at the later price.
Gemini 3.8 Flash at $0.75 / $3.75 until December 31
Google’s current fast model. Its price doubles to $1.50 / $7.50 on January 1, 2027.
For you: A cheap option today; check your bill in January.
Which Google model for which job
What to do now
Design so you can swap models: Argon may be the cheapest frontier option for a while, but its price will double and the date is unknown.
Get paid Gemini API access now so you are in the first wave. Meanwhile, test Gemini 3.8 Flash before its January price rise.
Argon reaches Google AI Ultra first. On other Google plans, nothing changes yet.