Short answer: Gemini 4 Argon is Google's new frontier AI model, introduced on September 30, 2026 and made broadly available on October 13. It is built for long, complex work: software engineering, deep document analysis, and multi-step tasks that run for a long time. It accepts text, images, audio, and video in a one million token context window and costs $2 per million input tokens and $10 per million output tokens at launch. Use it when a job needs sustained reasoning over a lot of mixed material. For the wider picture, see our guide to AI in business operations.
What is Gemini 4 Argon?
Argon is the first model in the Gemini 4 generation. Google dropped the familiar Pro and Flash labels for this generation and moved to element names. The API model ID is gemini-4-argon.
Google positions it for three kinds of work:
- Complex software engineering. Planning and executing changes across a large codebase.
- Enterprise knowledge work. Reading long reports, contracts, and data sets, then producing a usable answer.
- Long-horizon tasks. Jobs with many steps where the model has to keep its place for a long time.
We wrote about the pace behind this release in Google Says a New Model Every Two Weeks. Argon is that cadence turning into a product you can actually call.
What are the key specs?
| Spec | Gemini 4 Argon |
|---|---|
| Announced | September 30, 2026 |
| Broad availability | October 13, 2026 |
| API model ID | gemini-4-argon |
| Context window | About 1 million tokens |
| Input types | Text, images, audio, video |
| Output type | Text |
| Launch API price | $2 input and $10 output per 1M tokens |
What can a business do with it?
The one million token window with video and audio input is the practical difference. Most models read text well. Argon can take a recorded site walkthrough, a folder of photos, a voicemail, and a 200 page contract in one request and reason across all of it.
Real examples:
- A contractor uploads a walkthrough video and gets a scoped estimate draft with line items.
- An insurance agency feeds in a full policy set and asks where two carriers differ.
- A clinic turns recorded intake calls into structured records without a form.
- An agency audits a whole website export for content gaps in one pass.
How much does Gemini 4 Argon cost?
At launch the API price is $2 per million input tokens and $10 per million output tokens. A workload of 60 million input tokens and 10 million output tokens a month comes to $220. That puts it level with GPT-6.1 Sol on price and well under Claude Opus 5.5 at $4 and $20.
Two cautions. Video and audio consume tokens quickly, so a media-heavy workflow costs more than a text workflow of the same length. And launch pricing from any lab can change, so a platform should track cost per task, not assume today's rate holds.
How does Argon compare to other frontier models?
| Model | Best at | Input / output per 1M |
|---|---|---|
| Gemini 4 Argon | Long tasks over mixed media | $2 / $10 |
| GPT-6.1 Sol | Coding and agent workflows | $2 / $10 |
| Claude Opus 5.5 | Careful knowledge work and coding | $4 / $20 |
| Grok 4.7 | High-volume writing at low cost | $2 / $6 |
No single row wins every job. That is the point.
Should your business switch to Gemini 4 Argon?
Do not switch. Add. The businesses getting the most from AI right now are not loyal to one lab. Their platform sends video and long documents to Argon, routine drafting to a cheaper model, and the hardest reasoning to whichever model leads that week.
That only works if the model is a component and not the foundation. A platform built for native AI integration picks up a model like Argon when it ships, with no migration and no upgrade fee. Rented software makes you wait for a vendor's roadmap.
Frequently asked questions
When was Gemini 4 Argon released?
Google introduced Gemini 4 Argon on September 30, 2026 and made it broadly available on October 13, 2026, including through the Gemini API.
What is the Gemini 4 Argon context window?
Gemini 4 Argon has a context window of about one million tokens. It accepts text, images, audio, and video as input and returns text.
Is Gemini 4 Argon free?
Not at full strength. In the Gemini app it sits on Google's paid AI plans, and API use is billed per token at $2 input and $10 output per million tokens at launch.
Is Gemini 4 Argon better than GPT-6 or Claude Opus 5.5?
It depends on the job. Argon is strongest on long tasks that mix video, audio, images, and large documents. Other models lead on specific coding and reasoning benchmarks. Most businesses are best served by routing between them.
Why did Google stop using Pro and Flash names?
Gemini 4 uses element names for its models, starting with Argon. The names signal a new generation and separate roles, not a simple small, medium, large ladder.
Can Gemini 4 Argon run inside my own business software?
Yes, through the Gemini API. It needs a platform that holds your customer and project data, sends the right context with each request, and logs the cost of every call.
See Argon working on real business data
We have been building in AI since 2017, and every platform we ship is designed so a new model is good news the day it lands. Book a discovery call and we will show you frontier models running inside a platform built for your workflows.