Gemini 4 Argon Goes Global: Google's 1M-Token Flagship Opens to Every Developer — at Half Price Until 30 November

On 30 September, Google announced Gemini 4 Argon and gave access only to a vetted handful of cyber defenders. On 13 October, the door opened for everyone: developers, enterprises, researchers, and — for AI Ultra and Pro subscribers — ordinary users in the Gemini app. The rollout is global, the model ID is gemini-4-argon, and the introductory price is scheduled to expire in seven weeks.

What actually happened

The global rollout. Google's own announcement states that Argon is "now globally available today for developers and enterprises via the API in AI Studio and Gemini platform", for AI Ultra subscribers in Google Antigravity, and for AI Ultra and Pro subscribers in the Gemini app and AI Mode in Search (Google blog). The developer-facing post repeats the same for the paid tier of the Gemini API in AI Studio and Android Studio (developer announcement). On Google Cloud, Argon is live in Gemini for business across web, desktop and mobile, and in the Gemini platform under the model ID gemini-4-argon (Cloud blog).

The price has an expiry date. Argon launches at $2 per million input tokens and $10 per million output tokens through 30 November 2026, with cached input tokens at 95% off (Cloud blog). From 1 December the standard rate is $4 input and $20 output — double on both sides (Google blog). Two conditions matter for small buyers: Argon is not available on the Gemini API free tier, and not on the Google AI Pro plan (API docs). It is a paid-tier model, full stop.

One window for input and output. The model card and API documentation give a unified context limit of 1,048,576 tokens covering input, thinking and output together, with a default synchronous output cap of 65,536 tokens that long decode can extend up to the full 1M (model docs). Text, images, video, audio and PDF go in; text comes out. It is a sparse mixture-of-experts transformer trained on Google TPUs with JAX and Pathways, with a knowledge cutoff of March 2026 (model card).

Where it ranks, by the vendors' own numbers. Google reports Argon first on the Vals Index, which weights finance, coding, legal and tax work by each sector's contribution to US GDP (Vals Index); 77.9% on DeepSWE v1.1 for long-horizon software engineering; 51.3% on Zapier's AutomationBench for end-to-end execution across business functions; and 91.7% on LVBench for long-video understanding (Google blog). Treat those as vendor-reported benchmarks, not independent audits.

The cybersecurity angle is the unusual part. Google ties for first on CWE-bench v1 at 68% for remediating security vulnerabilities (Cloud blog), and — in its own model card — says it "cannot rule out" that Argon reached its cybersecurity and biological CBRN critical capability levels, which is why a guarded public release and an unguarded release to vetted defenders through the Fairwind Program run in parallel (model card). For enterprises, the headline safety number is different: on Gray Swan's indirect prompt-injection benchmark, attack success across 15 adaptive attempts drops to 0.7% (Cloud blog). That is the metric to watch if you plan to let an agent read your inbox or your documents.

Google's internal productivity claims. Teams using Argon in Antigravity increased agentic code submissions by more than 35% while cutting rollbacks by nearly a third; the Chrome team used it to finish an architectural refactor across hundreds of files in Chromium's 40-million-line codebase, described as months of work completed in weeks (Google blog). Google's Agent Development Kit team let Argon run overnight against a 200,000-line Python codebase and merged optimizations that cut in-memory session preparation time by 40% and cold imports by 21% (engineering deep dive). GitLab, Harvey, ServiceNow and Figma are quoted as early enterprise testers (Cloud blog).

Why it matters for Indian SMBs and mid-market teams

Three things, in the order they hit your budget.

First, the pricing window is a real buying lever. An Indian team processing long contracts, tender documents, or financial statements will burn output tokens on exactly the tasks Argon is built for — long, multi-step document work. Between now and 30 November the effective rate is half of what it becomes. If you have a bounded project (a contract-review pilot, a quarterly reporting automation), running it inside the window is cheaper than deferring it to December.

Second, the free tier is not a fallback. Teams already using free Gemini API quotas for experiments will find Argon simply absent. Plan the migration of any prototype you want on Argon around a paid-tier budget, and compare against your current model before switching — a flagship at $4 per million output tokens is a different financial instrument from the Flash-class models most SMB pilots run on.

Third, the prompt-injection number is the one to demand from every vendor, including us. Agents that read email, invoices, and web pages are exactly the SMB use case with the most automation upside and the most exposure. A 0.7% attack success rate across adaptive attempts is a meaningful bar; an agent that cannot state its injection-resistance posture should not be pointed at your mailboxes.

Where Guruji Corporation fits in

We build and run exactly this kind of automation for mid-market teams: document workflows, agent integrations, and the cost modelling that decides which model belongs on which task. Our practice is to benchmark a candidate model against your actual documents and your actual volume before anything goes live, then keep the routing under review when pricing changes — because on 1 December the arithmetic on Argon changes. If long-document or agent work is on your roadmap, this is the right month to prototype it.

Sources

Gemini 4 Argon Goes Global: Google's 1M-Token Flagship Opens to Every Developer — at Half Price Until 30 November

What actually happened

Why it matters for Indian SMBs and mid-market teams

Where Guruji Corporation fits in

Want to see how we can build your project?