Google has released Gemini 4 Argon, the first model in its Gemini 4 family and the one it calls its most capable yet. The company announced broad availability on Tuesday, Oct. 13, about two weeks after first teasing the model and giving it to a small group of cybersecurity partners.
Google says Argon is built for “deep work”: long, multi-step tasks in coding, finance, law, cybersecurity and science. Reuters framed the launch as Google’s bid to win back ground from OpenAI and Anthropic after months of delays and a leadership shakeup at its DeepMind lab.
Who Can Use It Right Now
The short answer: people who pay. Here is how access breaks down, according to Google:
- Gemini app and AI Mode in Search: Google AI Pro and Google AI Ultra subscribers.
- Google Antigravity (Google’s AI coding tool): Ultra subscribers only.
- Developers and businesses: the paid tier of the Gemini API, plus AI Studio, Android Studio and Gemini for business.
- Scientists: Google is offering 10,000 selected researchers six months of free access to Antigravity’s highest usage tier, subject to eligibility and capacity.
- Cyber defenders: Google is expanding its Fairwind Program, which gives vetted security teams a version of Argon without its cyber guardrails, to hundreds of additional partners.
Neither Google’s announcement nor early coverage says when, or whether, free Gemini users will get Argon. That is a notable contrast with OpenAI, which brought GPT-6 to free ChatGPT users last week.
What It Costs
For consumers, Engadget reported earlier this year that Google AI Pro costs $20 a month and that Ultra now starts at $100 a month, down from $250, with a $200 Ultra option for the heaviest users.
For developers, Google set introductory API pricing of $2 per million input tokens and $10 per million output tokens, with cached input 95% off. Those rates double, to $4 and $20, after Nov. 30. Anyone building on Argon should budget for the higher price now rather than be surprised in December.
What Google Says It Can Do
Google’s announcement leans heavily on benchmarks and internal results. Among the claims:
- It ranks first on the Vals Index, which tests finance, coding, legal and tax work.
- Google engineers using Argon in Antigravity submitted more than 35% more AI-generated code changes, while code rollbacks fell by nearly a third.
- It helped finish a Chrome architecture overhaul in weeks that had been months of work.
- On a prompt-injection benchmark run by the security firm Gray Swan, attacks succeeded only 0.7% of the time across 15 adaptive attempts.
- Google says the model hedges less, repeats itself less and follows style instructions better than earlier Gemini models.
Google also points to science projects: faster quantum chemistry software, five new enzyme designs that showed activity in the lab, and a rare-disease case in which Argon helped spot a chromosomal rearrangement that routine software had missed.
Safety Claims
Google says it tested Argon under its Frontier Safety Framework and added layered defenses. In external red-teaming on biology, testers could unlock less than 1% of the sub-tasks they tried, and violations on an adversarial test set fell by nearly half compared with earlier Gemini models. Google also says attempts to “cheat” on coding tests by taking shortcuts dropped to 2%, and that it runs an automated monitor to flag misaligned behavior in real time.
The Catch
Almost every number above comes from Google or its partners. Android Authority noted that real-world performance “remains to be seen,” and no independent lab results were part of the launch. Benchmarks also tend to measure narrow tasks; how a model handles your messy spreadsheet or half-finished code is a different test.
The coding claims deserve special scrutiny. More AI-written code is not automatically better code, and recent surveys have linked AI coding tools to outages at large software teams. Google’s own figure on fewer rollbacks is encouraging, but it comes from Google’s internal use.
Should You Pay for It?
- If you already pay for Google AI Pro or Ultra: try Argon on a task you know well, such as summarizing a long contract or debugging code, and compare it with what you use now.
- If you use free Gemini: there is no reason to upgrade on launch-day claims alone. Wait for independent tests and hands-on reviews over the next few weeks.
- If you build apps on the API: the introductory price makes this a good window to test, but plan for the doubled rate after Nov. 30.
- If you run a small business: treat Argon like any AI tool for legal, tax or financial work. Use it for drafts and analysis, and have a qualified person check anything you rely on.
More on Contoh
- Free Gemini Users Will Be Limited to Google’s Smallest Model Starting Oct. 9
- GPT-6 Comes to Free ChatGPT Today. Here’s What Changes, and What to Watch For
- AI Coding Tools Are Linked to Outages at Most Big Software Teams, Surveys Find
Sources
- Google: Gemini 4 Argon: Built to solve your hardest problems
- Slashdot (summarizing Reuters): Google Launches Flagship AI Model Gemini 4 Argon
- Android Authority: Gemini 4 Argon is finally here, and it’s not just for coders and developers
- TechCrunch: Google releases Gemini 4 Argon, called its most powerful model yet
- Engadget: The Google AI Ultra plan now starts at $100 a month