Atlas AI Resources

← Back to AI News

AI products

Google opened Gemini 4 Argon to paying users. Free access is still missing

Published October 15, 2026

On October 13, Google made Gemini 4 Argon—its newest frontier model (a top-tier AI system for hard, multi-step work)—available more widely after a limited cyber-defender preview in late September. Google’s blog calls it “available for everyone.” Ars Technica and The Verge report that free users still cannot use it; you need a paid Gemini plan or a paid API account.

What happened

Google DeepMind says Argon is live for developers and businesses through the Gemini API in Google AI Studio and the Gemini platform, for Google AI Ultra subscribers in Google Antigravity (Google’s agent-first coding environment), and for Ultra and Pro subscribers in the Gemini app and AI Mode in Search. Free access is not included. The Verge says it is unclear when—or if—Plus or free users will get Argon.

Ars Technica puts Pro at about $20 a month and Ultra starting at about $99 a month. In Antigravity and AI Studio, Argon is Ultra-only. Pro users can still use Argon for code in the Gemini app’s Canvas mode, but heavier multi-agent coding stays behind Ultra. Google is keeping Gemini 3.1 Pro available for people who need more than Flash models but cannot reach Argon yet.

For paid API use, Google lists $2 per million input tokens and $10 per million output tokens through November 30, 2026, with cached input tokens at 95% off; after that, Google says the price becomes $4 and $20 per million. Tokens are the small chunks of text a model reads and writes. The API also advertises a 1 million unified input-output token limit, so one request can carry a very long document or chat.

Google claims Argon ranks first on the Vals Index (a vendor-cited score across finance, coding, legal, and tax) and beats rivals such as GPT-6 Astra and Claude Opus 5.5 on several benchmarks. Those leads are Google’s reported results, not independent scores we verified here. Google also describes stronger refusal of dangerous requests on an adversarial set, an automated misalignment monitor that watches agent steps, and a leading Gray Swan prompt-injection result (0.7% attack success across 15 adaptive attempts, per Google). A green monitor or high confidence score is still not permission to take irreversible actions; Google says it trained Argon to pause for human confirmation first.

Why it matters to a normal person

If you already pay for Gemini Pro or Ultra, Argon is the upgrade path for harder writing, research, and coding inside Google’s apps. If you build software, the temporary API discount makes trials cheaper through late November—but plan for the price doubling after that, as Ars Technica notes. Google is also offering 10,000 scientists the highest Antigravity usage tier free for six months, subject to eligibility, capacity, and selection.

The bigger story is access. Frontier models are often announced weeks before ordinary users can touch them. This launch closes that gap for paying Google customers, while free ChatGPT-style access remains a separate product decision. Per-token price comparisons can mislead too: The Verge notes that a cheaper sticker rate does not always mean a cheaper finished task if one model uses more tokens than another.

What it does not mean

It does not mean Gemini 4 Argon is free for everyone, despite Google’s “available for everyone” headline. Free and lower-tier users are still on older or lighter models for now.

It also does not mean Google has proved Argon is the best model in the world. Benchmark leads and internal engineering anecdotes come from Google and early-access partners. Treat them as company-reported until outside labs publish their own numbers. Availability is not proven superiority, and a valid API answer is still not the same as a correct one without review.

Sources