Google Gemini 4 Is Finally Here: What Argon Can Do and Why Google Stock Moved

For most of 2026, one question kept following Google around: where is Gemini 4?
On Wednesday, September 30, Google finally answered. It showed off Gemini 4 Argon, the first model from its next Gemini generation. There is a twist, though. You can't use it yet. Unless you work at one of the few cybersecurity firms Google picked for early access, Argon is something you will read about before you get to try it.
Here is what we know so far: what the model does, what it costs, why it took so long, and what the news did to Google stock.
What Is Gemini 4 Argon?
Argon is Google's new flagship AI model and the first member of the Gemini 4 family. Google says it was built for long, complicated jobs. Think software engineering, finance, legal work and cybersecurity, where a task can run through many steps before it is done.
Google CEO Sundar Pichai introduced the model on X. He said there had been a lot of talk about Google's next model, and he wanted to give people an early look "as soon as possible." Tulsee Doshi, who leads Gemini products at Google DeepMind, told Axios that Argon is a well-rounded model with top-level skills across several areas.
If the name sounds familiar, that's because it leaked first. An early version had been showing up on the LMArena testing platform since mid-September, hiding under the label "gemini-3.8-flash." Argon was its internal codename. Now it is official.
Who Can Use Gemini 4 Right Now?
Very few people. Google is releasing Argon through something it calls the Fairwind Program. The first users are the US government and a selected group of cyber defenders. In that program, the model's job is to find, confirm and fix serious security holes in software, largely on its own.
Google is also part of the US government's voluntary pre-release testing process. The company says it will open Argon to developers, businesses and everyday users after more testing. Paying subscribers will get it first. No date has been shared.
This slow rollout makes more sense when you look at the rest of the industry. OpenAI recently cancelled its GPT-6.1 Astra release over security issues. Axios has also reported that OpenAI and Anthropic are looking into tens of thousands of cases of their AI agents misbehaving. Against that backdrop, Google going slow looks less like a delay and more like caution.
Gemini 4 Pricing and Key Specs
Detail | Gemini 4 Argon |
|---|---|
Input price | $2 per million tokens |
Output price | $10 per million tokens |
Output limit | 1 million tokens |
Built for | Coding, finance, legal work, cybersecurity |
Current access | Selected cybersecurity partners only |
For developers, the output limit is the number that stands out. It means Argon can work through a very long, multi-step task in a single pass instead of stopping halfway. Several market outlets have also read the pricing as Google trying to undercut its rivals.
How Gemini 4 Stacks Up Against OpenAI and Anthropic
Google shared results from 18 benchmarks and says Argon came out on top in 12 of them. It also claims the model beats OpenAI's GPT-6 Astra and Anthropic's Claude Fable on several coding and knowledge-work tests. Its biggest lead, according to Google, is in long-context testing, where the model has to keep track of a huge amount of information at once.
These numbers deserve a little caution. They are Google's own tests, picked and published by Google. The real verdict will come once outside developers start using Argon on their own work.
The Catch: Doubts From Inside Google
On the same day as the launch, Bloomberg reported that some Google employees were not fully convinced. According to its unnamed sources, Argon looked great on benchmarks but did less well when staff used it for actual tasks, especially coding. Some employees reportedly feel that Anthropic and OpenAI are improving faster. Others believe Gemini 4 has caught up.
Google disputed the report. It said it would be wrong to claim the model is underperforming at coding. One employee told Bloomberg there is broad agreement inside the company that Gemini 4 sits at the frontier of AI.
This is one of the oldest problems in AI. A model can be trained to score well on tests. Whether it holds up on messy, everyday work is something only real users can tell us.
Why Gemini 4 Took So Long
Gemini 4 arrives almost a year after Gemini 3, which is a long gap at the top of Google's lineup. Here is roughly how the year played out.
In May, Pichai said Google's next major model would arrive in June. Most people expected that to be Gemini 3.5 Pro. It never came out. Instead, Google kept shipping smaller and cheaper Flash models, including Gemini 3.6, 3.7 and 3.8 Flash. Gemini 3.8 Flash launched on September 2.
In July, Google confirmed it had started its biggest pre-training run yet for Gemini 4. Around the same time, Axios reported that low morale inside Google DeepMind was one reason for the delays.
August brought a leadership shake-up. Demis Hassabis stepped back from running Google DeepMind day to day, and Koray Kavukcuoglu took charge. Gemini's two co-leads also left the same day. Gemini 4 became Kavukcuoglu's first big launch.
In late September, Kavukcuoglu said at The Information's AI Agenda Live event that Gemini 4 was in post-training. He said Google wanted to release it much earlier than the end of the year. About a week later, Argon was out.
Google Stock Reaction: What Happened to GOOG and GOOGL
September 30 was a busy day for Google stock. During regular trading, shares bounced off their 200-day moving average of about $337. Analysts at Piper Sandler also said that TSMC, which makes Google's TPU AI chips, is sold out through the first half of 2028. That was a sign that demand for AI hardware is still strong.
GOOGL was up more than 2% at one point. Then the Bloomberg report came out, and the stock gave back most of that gain, closing up around 0.5%. After the market closed, Google announced Argon. GOOG then rose about 1.7% in after-hours trading, to roughly $346.52.
A few numbers help put this in context:
Alphabet shares were up about 42.6% over the past 12 months.
The company's market value sits at roughly $4.1 to $4.2 trillion.
GOOGL's 52-week high is $402.62, so the stock is still well below its peak.
GOOG's trailing P/E ratio is about 17, lower than its five-year median of about 24.
The market's reaction looked more like relief than excitement. Investors had spent months worrying that Google was falling behind in AI. Argon does not settle that debate, but it gives Google something real to point to.
✅ Pros
Google says Argon leads on 12 of the 18 benchmarks it shared.
Aggressive pricing: $2 per million input tokens, $10 per million output tokens.
A 1 million token output limit for long, multi-step work.
Built for heavy tasks in coding, finance, legal and cybersecurity.
A careful, safety-first rollout with government testing.
Another advantage
❌ Cons
AdNot open to the public yet, and no release date.
Early access is limited to selected cybersecurity partners.
A Bloomberg report says some Googlers found it weaker on real work than on benchmarks.
Benchmark claims are Google's own and not yet independently tested.d a con point here
Another disadvantage
GOOG vs GOOGL: Which Google Stock Is Which?
If you searched for "Google stock," you probably saw two tickers. Both belong to the same company, Alphabet.
GOOGL is Class A stock. Each share comes with one vote.
GOOG is Class C stock. These shares have no voting rights.
The two usually trade at very similar prices, and both moved on the Gemini 4 news. For most retail investors, the difference only matters if voting rights are important to you.
What Comes Next for Gemini 4
The wider release is the real test. A few things are worth watching over the coming weeks:
When Google opens Argon to paying subscribers, and at what price.
What independent testers find once they can run their own benchmarks.
How fast Gemini 4 reaches Google's everyday products. Versions of Gemini already power almost everything Google sells, from AI answers in Search to Maps, Gmail and Chrome.
Google needed a strong answer to OpenAI and Anthropic this year. Argon looks like one on paper. Now it has to prove itself in the hands of real users.
Disclaimer: This article is for information only and is not investment advice. Stock figures are as of September 30, 2026.
By Vikash Kumar, VSURGE MEDIA Kolkata
Ready to grow your business?
Let's discuss how Vsurge Media can help you achieve your goals.
Book a Free Consultation