Last updated: October 1, 2026
The quick answer
Gemini 4 Argon is Google’s new top-tier AI model, announced on September 30, 2026, and aimed at long, complex tasks like coding, research and cybersecurity. Right now only selected security partners can use it; Google says paid API customers and Google AI Ultra subscribers come next, but it has not given a date. If you use the free Gemini app, nothing changes for you yet.
What happened
Google introduced Gemini 4 Argon as the first model in its Gemini 4 generation. The company describes it as built for deep reasoning over long, multi-step workflows, with particular strengths in software engineering, enterprise knowledge work and finding and fixing security vulnerabilities.
Unusually, Google is not opening it to the public on day one. Instead, Argon is rolling out first to trusted cyber defenders through what Google calls the Fairwind Program. TechCrunch reports that the model can autonomously find, validate and patch serious software flaws, which is why Google is starting with security teams. 9to5Google adds that these vetted professionals can use it without some of the cyber guardrails applied to regular users.
Google also says it is taking part in the U.S. government’s voluntary process for pre-release model access, and that Argon is its most resilient model yet against indirect prompt injection, a trick where hidden instructions in web pages or documents try to hijack an AI assistant.
Key details at a glance
| Item | What Google has announced |
|---|---|
| Available now | Trusted cyber defenders via the Fairwind Program |
| Next in line | Paid Gemini API customers and Google AI Ultra subscribers, no date given |
| Launch API price | $2 per million input tokens, $10 per million output tokens (introductory) |
| Standard API price | $4 per million input tokens, $20 per million output tokens |
| Cached input | 95% discount on the input price |
| Output limit | Up to 1 million tokens, up from 64,000 |
On performance, Google says Argon scored ahead of OpenAI’s GPT-6 Astra and Anthropic’s leading models on several benchmarks, citing the independent Vals AI index, according to TechCrunch. 9to5Google lists scores such as 77.9% on the DeepSWE v1.1 coding test and 91.7% on the LVBench video-understanding test. As with any launch, these are early results that independent testing still needs to confirm.
What changes for you
For everyday Gemini users, nothing changes today. The model is not in the free Gemini app, and Google has not said when or whether it will reach lower-priced plans.
The bigger story is the direction. A much higher output limit means the model can produce very long results in one go, such as an entire code migration or a lengthy report. And by launching first with security teams, Google is signaling that the most capable AI models are now treated as dual-use tools whose release needs to be staged.
For developers, the introductory price matches what OpenAI charges for its cheaper GPT-6.1 Sol model, which we covered in our GPT-6.1 Sol explainer. The standard price, after the introductory period, is double that.
How to get ready
- Check your plan. Only Google AI Ultra is named for early consumer access. Do not upgrade just for Argon until Google confirms a date.
- Developers: watch the Gemini API model list and pricing page. Budget for the standard rate, not only the introductory one.
- Security teams: ask your Google Cloud contact about the Fairwind Program if you want early access.
- Everyone: keep reviewing AI output. Strong benchmark scores do not remove the need to check code, facts and sources.
Who it affects and who it doesn’t
- Affects now: security organizations admitted to Google’s Fairwind Program.
- Affects soon: paid Gemini API customers and Google AI Ultra subscribers, once the rollout expands.
- Doesn’t affect yet: free Gemini users and other paid plans.
Gemini is changing in other ways this fall too. See how Gemini Gems are becoming skills.
FAQ
Can I use Gemini 4 Argon in the Gemini app today?
No. At launch it is limited to selected security partners. Google says AI Ultra subscribers will be among the first consumers to get it, without a date.
Why is Google limiting the release?
Google highlights Argon’s ability to find and patch software vulnerabilities. Skills like that can also help attackers, so the company is starting with vetted defenders before a wider rollout.
Is it better than ChatGPT?
Google says Argon beats OpenAI’s GPT-6 Astra on several benchmarks, but few people can test it yet. Wait for independent reviews before deciding.




Leave a Reply