What is Google Argon?
Google Argon is the nickname most people use for Gemini 4 Argon, a new AI model from Google DeepMind. Google announced it on September 30, 2026, and opened it to the public on October 13. It's the first model in the Gemini 4 family, and Google calls it a frontier model, which is AI-speak for one of the most capable models around. It's built for three kinds of heavy lifting: real-world coding, business knowledge work and cyber defense.
What about that name? Past Gemini models were sorted into tiers like Pro and Flash. This one skips the tier and goes by a single word, Argon, which reporters are treating as the start of a fresh naming pattern. Why Argon? Google hasn't said.
What it can do, in everyday terms
If a typical AI chatbot is a helpful friend who answers questions, Google Argon AI is closer to a team of specialists who read the whole binder before they say a word.
- It can write a book-length answer, and then some. Earlier Gemini models topped out at about the length of a short novel in one reply. Argon can write roughly 750,000 words at once, several long books' worth. That's handy when the job is huge, like rewriting a big app or drafting a very long report.
- It can read a whole bookshelf. Google hasn't said exactly how much Argon can take in at once. But in one of Google's tests it worked through inputs of up to a million tokens, about a shelf of books, and kept track of the details better than its rivals did.
- It understands pictures and video. You can show it images and videos, including charts and documents full of visuals, and it can explain what's in them. On a test about understanding long videos, Google reports a score of 91.7%.
- It sticks with long jobs. Google built it for work with many steps: moving an old app to newer technology, digging through piles of financial or legal documents, or checking software for security holes.
What comes out is always text. Code, answers, reports and plans, yes. Pictures and videos, no. More on that below.
The three numbers that matter
1 million. AI models count text in tokens, small chunks of text that are often pieces of words. Argon can write up to a million of them in one answer, about 15 times the old limit of 64,000. For most people this simply means Argon can finish a giant task in one go instead of stopping halfway.
13 of 19. Google compared Argon with top models from OpenAI and Anthropic on 19 tests. Argon came first on 13, tied on one and trailed on five. These are Google's own launch-day numbers. One independent ranking, the Artificial Analysis Intelligence Index, puts Argon a bit behind the newest ChatGPT and Claude models overall. We break it down in Google Argon vs ChatGPT vs Claude.
0.7%. One sneaky way to attack an AI is to hide instructions inside a web page or email it reads. On Gray Swan's prompt-injection test, where attackers got 15 tries, those tricks worked on Argon 0.7% of the time, the lowest in Google's chart. Not zero, but low. Our Google Argon safety guide explains why this matters.
Can you use it, and is it free?
Yes, you can use it now. Google announced Argon on September 30, 2026, and made it publicly available on October 13, 2026. It's built for paying users, though.
- Everyday users: Google AI Pro or Ultra. If you subscribe to Google AI Pro or Google AI Ultra, you can use Argon in the Gemini app and in AI Mode in Google Search. Google AI Ultra subscribers also get it in Google Antigravity.
- Developers: the paid Gemini API. You can build with it through Google AI Studio and Android Studio, using the model ID
gemini-4-argon. It's also on Vertex AI, which Google now calls Gemini Enterprise Agent Platform. - Security teams: Fairwind. Google's program for vetted cyber defenders got Argon first, and on October 13 it grew to hundreds more partners.
Is Google Argon free? No. Google says Argon is not available on the Gemini API's Free tier, and it hasn't announced free access in the Gemini app. The free Gemini app still runs Google's other models.
For developers, the price is $2 per million tokens going in and $10 per million coming out through November 30, 2026. From December 1 it goes to $4 and $20. Cached input is 95% off. That's a developer price, not a monthly plan for regular people.
Newer news: Gemini 4 Carbon. On October 28, Google announced Gemini 4 Carbon, which it calls its strongest model. For now it goes only to selected cyber defenders, and Google says everyone else gets it as soon as possible. We cover it in our Gemini 4 Carbon guide.
Who is it really for?
Google's rollout order tells you who it built Argon for.
- Security teams. Argon is unusually good at finding flaws in software. Google gave defenders access first, so they can find and fix problems early.
- Developers. Coding is one of its three main jobs, and a million-token answer means big code changes in one pass. Developers: see the full technical guide on gemini4argon.com.
- Businesses. Some of Argon's biggest leads in Google's table are on finance, legal and office-automation tasks, the kind of work that lives in documents and spreadsheets.
- Everyone else. For a quick recipe or a birthday message, you won't need a model this big. You're more likely to feel Argon through the apps and tools built on top of it.
What it can't do: make images or videos
This is the big one for creators and marketers. Google Argon writes. It doesn't draw, film or sing. It takes text, pictures and video in, and gives text back. Google hasn't said it makes images, video or audio.
The good news is that Google already has separate models for making things, and you don't have to wait for any of them:
- Veo 3.1 makes short videos with sound built in.
- Gemini Omni makes and edits short clips while you chat with it.
- Nano Banana 2 makes and edits images.
All three are on Lumeta, next to other top video models like Seedance and Kling, in one account. Lumeta is made by the team behind this site.
New to all this? Our guide on how to try Google's best AI today walks you through your first video, step by step.
Should you care?
Argon is big news for the people who fix software, and a nice-to-have for everyone else.
If you work in security, yes. If you build software or run a business that lives in documents, the paid API and Google AI Pro or Ultra are worth a look. If you mostly use AI for everyday stuff, there's no rush: the free Gemini app still works fine. And if you want to make pictures or video, Argon was never the tool for that, and the right ones are ready now.
Google Argon questions, answered
Is Google Argon real?
Yes. Google announced Gemini 4 Argon on September 30, 2026, and made it publicly available on October 13, 2026. Many people shorten the name to Google Argon.
What does Argon mean?
Nobody outside Google knows yet, because Google hasn't given a reason. Reporters see it as a fresh way of naming Gemini models, in place of tiers like Pro and Flash. As a fun aside, argon is also a chemical element: a noble gas, number 18 on the periodic table, whose name comes from the Greek word for idle. That's a coincidence we enjoy, not Google's reason.
Is Google Argon better than ChatGPT?
On Google's own test table, Argon beat GPT-6 Astra, a model from OpenAI (the company behind ChatGPT), on most tests but not all. These are Google's numbers, independent results aren't out yet, and Google compared models, not the ChatGPT app itself.
Can I use Google Argon now?
Yes, if you pay. Google AI Pro and Google AI Ultra subscribers can use it in the Gemini app and in AI Mode in Search. Developers can use it on the paid Gemini API. Google AI Ultra subscribers also get it in Google Antigravity.
Is Google Argon free?
No. Argon is for Google AI Pro and Ultra subscribers and for developers on the paid API. Google says it is not available on the API's Free tier, and it hasn't announced free access for regular users. The free Gemini app runs Google's other models.
What is Gemini 4 Carbon, and is it better than Argon?
Gemini 4 Carbon is the model Google announced on October 28, 2026, and calls its strongest. In Google's own table it beats Argon on all 17 tests and costs the same. For now it goes only to selected cyber defenders. Read our Gemini 4 Carbon guide for the details.
Can Google Argon make pictures or videos?
Google hasn't said it can. Argon reads text, images and video, but it answers in text. To make videos, Google has Veo 3.1 and Gemini Omni, and for images it has Nano Banana 2. You can use all three today, including on Lumeta.
Sources: Google's announcement, 9to5Google, The Next Web, Gemini plans, Google DeepMind: Veo, Royal Society of Chemistry: argon.