Faysal AzizGemini 4 Argon goes global: check access before you compare scores
Google’s Gemini 4 Argon is now rolling out worldwide, according to THE DECODER. The question for most people is less “did it win the AI race?” and more “can I use it in the place I work, at a price I can live with?” The answers differ for API developers, paid subscribers and free users.
By George the bot
Edited and approved by Faysal Aziz
Published
Which route gets Argon?
THE DECODER reports that developers can reach Argon through Google AI Studio’s API and the Gemini platform. Google AI Ultra subscribers can use it in Antigravity. AI Ultra and Pro subscribers also get it in the Gemini app and Search’s AI mode. Free Gemini users and Plus subscribers remain on Flash models under the new tier structure. A global rollout, then, does not mean every account sees the same model in every product.
If you are considering a switch, check the exact product and plan you use, not just the model name in an announcement. API access, an app subscription and access inside a coding tool are separate buying decisions. Limits can change during a rollout, so verify the current model picker and account terms before promising a team a feature.
The introductory price is not the enduring price
At launch, THE DECODER lists promotional API prices of $2 per million input tokens and $10 per million output tokens. Its comparison gives the regular rates as $4 and $20. Google also says cached inputs receive a 95% discount; the article derives a promotional cache-read figure from that claim rather than quoting a separately stated Google price. Workloads with many generated tokens or little reusable input will have a different bill from those that benefit heavily from caching. Price a representative workload, and use the regular rate when planning beyond a promotion.
Good demos, narrower evidence
Argon leads the Vals Index cited by THE DECODER, but the article notes that other independent overall rankings put it behind a competitor. A leaderboard depends on its test mix and date; no single position tells you how it will perform on your own prompts, documents or code. Google also reports internal engineering gains, including more agent-driven code submissions and fewer rollbacks, plus faster work on a Chromium refactor. Those are Google's accounts, not a controlled study showing the same gains for every team.
The practical next step is a small evaluation. Pick tasks you actually do, decide in advance what a correct answer or accepted patch looks like, and compare quality, time and total cost against the model you already use. Include failure cases and any privacy or integration requirements. If you are just trying Argon in an app, make sure your plan includes it first. The rollout is useful news; the buying decision still belongs to your workload.