Gemini 4 Argon, Google’s newest frontier AI model, is now globally available — at half price, but only until November 30. After a limited September 30 debut for cybersecurity teams, Argon opened to the world today, October 13, via the paid Gemini API, AI Studio, Antigravity, the Gemini app, and AI Mode in Search. Intro pricing: $2 per million input tokens and $10 per million output, doubling to $4/$20 on December 1.
The news comes from Google’s official launch post by Koray Kavukcuoglu, SVP of Google DeepMind and Chief AI Architect. Argon is Google’s first flagship pro model in over seven months, built for long, complex, multi-step work — the deep work that commands premium freelance rates.
What launched today, and who gets it
Argon rolls out globally today across five surfaces: the paid Gemini API in AI Studio and Android Studio (model ID gemini-4-argon); Gemini for business and Antigravity for enterprises; the Gemini app and AI Mode in Search for Google AI Ultra and Pro subscribers (Antigravity needs Ultra); a free 6-month Antigravity top tier for 10,000 scientists; and an expanded Fairwind Program for cyber defenders.
Note the gaps: no free-tier access, and Antigravity access requires Ultra — Google is clearly using Argon to pull power users up the pricing ladder.
The price window: $2/$10 until November 30, then it doubles
The introductory API pricing — $2 per million input tokens, $10 per million output tokens, cached input 95% off — matches Claude Sonnet 5.5 and undercuts Claude Opus 5.5 by half. After November 30, 2026, it becomes $4/$20, exactly Opus 5.5’s tier (GPT-6 Astra sits at $10/$50).
That gives builders a defined seven-week window to prototype and ship on frontier-class intelligence at mid-tier prices. If your unit economics work at $2/$10 but not at $4/$20, the calendar is your business plan: validate before December, or design for the higher rate from day one.
The spec that matters: 1 million output tokens
Argon’s headline feature is an industry-leading 1-million-token unified input-output limit — up from 64,000 tokens of output in previous Gemini generations. Google points developers to a new continuation protocol in the docs to manage long generations. The model takes text, images, video, audio, and PDFs as input and produces text output.
Why it matters for money: one continuous generation can produce an entire codebase refactor, a full legal brief, or a book-length report without the orchestration tax of chaining dozens of calls. Fewer calls, fewer failure points, simpler products.
The evidence: what Google says Argon already did
Google’s launch post is unusually specific about internal results — all vendor-reported, so treat them as claims to verify:
- #1 on the Vals Index (68.9%), measuring finance, coding, legal, and tax work — ahead of Claude Sonnet 5.5 (67.0%) and Claude Opus 5.5.
- Agentic coding: Google teams using Argon in Antigravity increased agentic code submissions by more than 35% while reducing rollbacks by nearly a third.
- Chrome refactor: Argon helped untangle a ~6-million-line legacy folder inside Chromium’s 40-million-line codebase across hundreds of files — “months of work in weeks.”
- Real vulnerability: Wiz used Argon in its Scan for Good initiative to find and fix a file-download bypass that exposed a master key in healthcare software — protecting hundreds of thousands of patient records.
- Speed: 2.85x inference speedup on Gemma 4 E2B via vLLM on an A100 (~37,900 to 108,000+ tokens/second) with zero quality drop.
- Safety: 0.7% attack success on the Gray Swan prompt-injection benchmark across 15 adaptive attempts; agent shortcut attempts down to 2%.
The money angle: how to profit from the Argon window
Three plays. Build on the intro price, ship before December: seven weeks of frontier intelligence at $2/$10 is a subsidized R&D budget — prototype an agent service (contract review, codebase audits, document processing), validate willingness to pay, and price against the $4/$20 standard rate so margins survive December 1. Sell outcomes, not hours: a million output tokens means one generation can deliver a complete migration or full audit as a single artifact — outcome pricing beats hourly by multiples, and the capability gap is temporary. Borrow enterprise credibility: with Deloitte, LexisNexis, and Ambience Healthcare testing Argon, selling Argon-powered services to mid-market firms gets dramatically easier.
What to watch next
Independent benchmarks against GPT-6 Astra and Claude Opus 5.5 (Google’s numbers are self-reported), the demand impact of the December 1 price doubling, and whether Argon ever reaches cheaper plans. The specs are real, the price window is real, and the independent verdict is still pending.
FAQ
Is Gemini 4 Argon available now?
Yes. Since October 13, 2026, it is globally available via the paid Gemini API, AI Studio, Antigravity (Ultra), and the Gemini app and AI Mode in Search (Ultra and Pro).
How much does Gemini 4 Argon cost?
Introductory API pricing is $2 per million input tokens and $10 per million output tokens, with cached input 95% off, through November 30, 2026. From December 1, the price doubles to $4/$20.
Can free Gemini users access Argon?
No. Argon is not on the free tier or free API. In the Gemini app you need Google AI Pro or Ultra; Antigravity requires Ultra.
What is Argon’s context and output limit?
A unified 1,048,576-token window shared between input and output, with a continuation protocol for long generations. It accepts text, images, video, audio, and PDFs, and outputs text.
How does Argon compare to GPT-6 Astra and Claude Opus 5.5?
Google claims #1 on the Vals Index and leads on several coding benchmarks, but these are vendor-reported. Independent third-party benchmarks are still pending — test against your own workloads before committing.
Sources
- Gemini 4 Argon: Built to solve your hardest problems — official Google blog, Koray Kavukcuoglu, October 13, 2026 (availability, pricing, benchmarks, internal results)
- Google launches flagship AI model Gemini 4 Argon — Reuters, October 13, 2026