Google has a new flagship AI model, and it arrived with a bang. On October 13, 2026, the company released Gemini 4 Argon to developers and businesses worldwide — a model Google says is built for “deep work” on the complex, multi-step tasks that define serious professional use: coding, finance, law, cybersecurity, and scientific research.
The launch lands at a dramatic moment for Google DeepMind. It follows months of delays, the September cancellation of Gemini 3.5 Pro, and a leadership shakeup after Demis Hassabis stepped aside. In other words, Argon isn’t just another model release — it’s Google’s answer to everyone who asked whether it could still lead the AI race.
Here is everything we know about Gemini 4 Argon so far: what it can do, how much it costs, where you can access it, and whether it should matter to you.
What Is Gemini 4 Argon?
Gemini 4 Argon is Google’s new flagship AI model — the top tier of the Gemini family, replacing whatever came before it as the model the company puts forward for its hardest problems. It is generally available now through the paid Gemini API (model id gemini-4-argon) and Google AI Studio, meaning developers and businesses anywhere can start building with it today rather than waiting on a waitlist.
Google’s framing is specific: Argon is designed for “deep work.” That’s shorthand for long, complex, multi-step tasks where an AI has to hold context, plan across stages, and produce work that professionals can actually trust — writing and reviewing large codebases, digging through financial documents, parsing legal language, hunting security vulnerabilities, or assisting with scientific analysis. This is the territory where flagship models justify their existence, and Google is aiming Argon squarely at it.
The name itself is a statement. Argon — an inert noble gas — suggests something stable, reliable, and unreactive: exactly the reputation Google wants after a rocky stretch of delayed and cancelled releases. Whether the model lives up to that promise will be decided by real-world testing in the weeks ahead.
Key Capabilities
Google is positioning Argon as a specialist in professional-grade work rather than a general chit-chat upgrade. Here’s where it’s supposed to shine.
Coding
Coding is Argon’s headline strength. Google says teams using the model inside Antigravity — its agentic coding environment — increased agentic code submissions by more than 35% and reduced code rollbacks by nearly a third. That’s a striking claim: more code shipped, fewer reversals. Agentic coding is the current frontier, where AI doesn’t just autocomplete lines but plans, writes, tests, and revises whole features on its own. If Argon’s numbers hold up in independent testing, it becomes an immediate contender for developers choosing a flagship coding model.
Finance and Law
Finance and law share a common demand: precision over long, dense documents. Google lists both as core Argon strengths, and the early tester roster backs that up — Deloitte and LexisNexis are among the companies that have been trying the model. LexisNexis is one of the giants of legal research, and Deloitte is a global professional-services firm; their involvement suggests Argon is being evaluated for the kind of document-heavy analytical work where mistakes are expensive. As always, treat vendor claims about accuracy as a starting point for your own testing, not a guarantee.
Science and Cybersecurity
On the science side, Google is putting real resources behind the pitch: the company will offer 10,000 scientists six months of free access to Antigravity’s highest usage tier. That’s a meaningful signal about where Google thinks Argon adds value — research workflows that combine code, data analysis, and literature review.
Cybersecurity gets its own dedicated track through the Fairwind Program, which gives vetted cybersecurity defenders a version of the model without the usual cyber guardrails, so they can use it for legitimate defensive work like threat analysis and vulnerability research. Google says hundreds of new partners are joining the program. It’s a notable move: the same capabilities that make a model useful for defense could be misused for offense, so restricting the unguarded version to vetted defenders is Google’s way of threading that needle.
Early testers named by Google also include Ambience Healthcare and Isomorphic Labs, spanning healthcare AI and drug discovery — further evidence that Google is pitching Argon as a model for industries where the stakes are high and the tasks are hard.

Gemini 4 Argon Pricing
Argon is a premium model and it’s priced like one. The launch API pricing is:
| Token type | Price per 1M tokens (launch rate) | Price per 1M tokens (from Dec 1) |
|---|---|---|
| Input tokens | $2 | $4 |
| Output tokens | $10 | $20 |
Two things to note. First, output tokens cost five times input tokens — typical for flagship models, but it means heavy generation workloads add up fast. Second, these launch rates double after November 30: standard pricing applies from December 1, taking input to $4 and output to $20 per million tokens. If you’re planning a project around Argon, that deadline matters.
Also note: Gemini 4 Argon is not available on the API free tier. You’ll need a paid Gemini API setup to use it programmatically.
Check current pricing on Google’s official pricing page before committing to anything — rates and tiers can change, and the December 1 increase is already scheduled.
Where You Can Access It
Google is rolling Argon out across its developer and consumer surfaces at once. Here’s the full list:
| Where | Who gets it | What it’s for |
|---|---|---|
| Gemini API (paid) | Developers and businesses | Building apps and integrations with model id gemini-4-argon |
| Google AI Studio | Developers | Prototyping and testing prompts |
| Antigravity | Google AI Ultra plan subscribers | Agentic coding environment |
| Gemini Enterprise | Eligible enterprise subscriptions | Business and team deployments |
| Gemini app | Google AI Pro and Ultra subscribers | Everyday chat and assistant use |
| AI Mode in Search | Google AI Pro and Ultra subscribers | AI-powered search experiences |
| Android Studio | Developers | AI-assisted Android development |
| Firebase AI Logic | Developers | AI features inside Firebase apps |
The pattern is clear: Argon is gated behind paid tiers everywhere. Free-tier users won’t see it in the API, and the consumer surfaces (Gemini app, AI Mode in Search) require a Pro or Ultra subscription. Google is treating this as a premium product from day one.
How to Try It
Want to get your hands on Argon? Here are the two fastest paths.
Option 1: Try it in the Gemini app
- Check your subscription. Argon in the Gemini app requires a Google AI Pro or Ultra subscription — sign up if you don’t have one.
- Open the Gemini app on the web or your phone and sign in with that Google account.
- Select the model. Look for Gemini 4 Argon in the model picker (availability may roll out progressively).
- Test it on hard tasks. Give it something genuinely complex — a multi-file coding problem, a dense document to analyze — since that’s what the model is built for.
Option 2: Use it via the API or AI Studio
- Set up billing. Argon is not on the API free tier, so enable billing on your Google Cloud project or paid Gemini API account.
- Open Google AI Studio and select the
gemini-4-argonmodel to prototype prompts in the browser. - Call the API using the model id
gemini-4-argonin your existing Gemini API code — the endpoint works the same as other Gemini models. - Watch your token usage. Remember the pricing: $2 input / $10 output per million tokens now, doubling after November 30. Set budget alerts before you run large jobs.

The Rocky Road Here
Argon’s launch can’t be understood without the turbulence behind it. Google’s flagship-model timeline has been bumpy: Gemini 3.5 Pro, which had been promised for June, was cancelled outright in September — an unusually public admission that something wasn’t ready. Around the same time, DeepMind went through a leadership overhaul after Demis Hassabis stepped aside.
That context cuts two ways. On one hand, it explains why Google is leaning so hard on early results and enterprise testers — it needs Argon to land. On the other hand, a cancelled predecessor and reshuffled leadership are exactly the conditions under which a company rushes. The fair read: Argon’s claimed capabilities are impressive, but independent benchmarks and real-world developer reports in the coming weeks will tell the real story. If you’re evaluating it for production use, run your own evaluations rather than trusting launch-day numbers alone.
Early Results
Google’s launch-day evidence rests on two pillars: internal coding metrics and named enterprise testers.
The coding numbers are the strongest claim. Teams using Argon inside Antigravity — Google’s agentic coding environment — reportedly increased agentic code submissions by more than 35% while reducing code rollbacks by nearly a third. If those figures generalize, they describe a model that doesn’t just write more code but writes better code, which is the harder and more valuable trick.
The tester roster is designed to impress: Deloitte (professional services), LexisNexis (legal research), Ambience Healthcare (healthcare AI), and Isomorphic Labs (drug discovery, and notably a DeepMind sibling). These are serious organizations running serious evaluations. But note what we haven’t seen yet: independent third-party benchmarks, head-to-head comparisons against rival flagships, and unfiltered developer feedback. Those will arrive soon — and they’ll matter more than launch-day marketing.
For context on how the AI assistant landscape looked before this launch, see our ChatGPT vs Claude vs Gemini comparison and our roundup of the best AI chatbots compared.
Verdict: Who Should Care
Gemini 4 Argon is clearly aimed at a specific audience — and it’s probably not casual users, at least not yet.
- Developers and engineering teams should pay the closest attention. The agentic coding claims, the Antigravity integration, and availability in Android Studio and Firebase AI Logic make this a direct pitch to people who ship software. If the 35% submission boost holds up, that’s a productivity story worth testing.
- Businesses in finance, law, and healthcare should watch the Deloitte, LexisNexis, and Ambience Healthcare evaluations. These are the industries where “deep work” capabilities translate into real ROI — and real risk if the model hallucinates.
- Security teams should look at the Fairwind Program, which offers vetted defenders an unguarded version for defensive work.
- Researchers get a genuine free on-ramp: six months of Antigravity’s top tier for 10,000 scientists.
- Students and casual users can probably wait. The consumer surfaces require paid subscriptions, and our guides to the best free AI tools for students and ChatGPT vs Gemini for students cover what you can do without paying.
The bottom line: Argon is Google’s biggest swing since the Gemini era began, launched under pressure and priced as a premium product. The claims are bold, the early testers are credible, and the pricing clock is ticking toward December 1. Now it’s up to independent testing to decide whether Argon is the comeback Google needs — or just another launch-day story.
Frequently Asked Questions
When was Gemini 4 Argon released?
Google released Gemini 4 Argon on October 13, 2026, to developers and businesses worldwide. It launched as generally available through the paid Gemini API (model id gemini-4-argon) and Google AI Studio, alongside availability across Antigravity, Gemini Enterprise, the Gemini app, AI Mode in Search, Android Studio, and Firebase AI Logic.
How much does Gemini 4 Argon cost?
The launch API pricing is $2 per million input tokens and $10 per million output tokens. These rates double after November 30 — standard pricing from December 1 is $4 input and $20 output per million tokens. It is not available on the API free tier. Check Google’s official pricing page for current rates, since pricing can change.
Is Gemini 4 Argon free to use?
No. There is no free-tier access: the API requires a paid setup, Antigravity access needs a Google AI Ultra plan, and the Gemini app and AI Mode in Search require Pro or Ultra subscriptions. The main exception is Google’s offer of six months of free Antigravity top-tier access to 10,000 scientists.
What is the Fairwind Program?
Fairwind is Google’s program giving vetted cybersecurity defenders access to a version of Gemini 4 Argon without the usual cyber-related guardrails, so they can use it for legitimate defensive security work. Google says hundreds of new partners are joining the program.
What happened to Gemini 3.5 Pro?
Gemini 3.5 Pro was cancelled in September 2026 after being promised for June. The cancellation, along with a DeepMind leadership overhaul after Demis Hassabis stepped aside, preceded the Gemini 4 Argon launch — making Argon Google’s attempt to reset its flagship-model story.