Bloomberg/Getty Images
- Google is rolling out its new frontier model, Gemini 4 Argon, to paying users.
- Google says Argon excels in coding, as well as knowledge work in finance and law.
- It's Google's opportunity to reassert itself in the frontier race.
Google says it's back on the AI frontier, and now everyone can put that claim to the test.
The company announced on Tuesday that its new Gemini 4 model, dubbed "Argon," is rolling out to the public and will be available to Gemini app users and developers.
Google is positioning Argon as its new workhorse model, stating that it excels at coding, cybersecurity, and knowledge work in finance and law.
Tulsee Doshi, Gemini product lead at Google DeepMind, called Argon a "powerhouse" model and said it would be the first of a family of new Gemini models rolling out over time.
As is often the case with the most advanced AI models, Argon is available only to paying users. It's rolling out to Pro and Ultra Gemini app subscribers, meaning users will have to pay at least $19.99 per month to access it. Those subscribers will also see Argon in AI Mode while searching on Google. Developers will also need to pay for the new model.
Google first unveiled Argon last month, but it was only available to a set of trusted partners to proactively test it and address cyber vulnerabilities.
The new model arrives amid heated competition between the big AI companies. Meta said on Monday that its new Muse Wonder class of models would bring the company to the frontier, the industry term for the most advanced AI, while also powering its buzzy personal agent, Muse. Last month, OpenAI started rolling out its latest family of frontier models, GPT-6.
Argon gives Google an opportunity to reassert itself in the AI race after the company chose not to launch its previously-promised frontier model, Gemini 3.5 Pro.
Cornhole-playing robot
Google has been testing its new Gemini 4 model internally with employees for several weeks and said it had helped engineers increase the production of AI-generated code by 35%, while helping improve other Google products such as Chrome.
"I actually think it's been one of the biggest game changers for us," Doshi told Business Insider, referring to the increased amount of internal testing for Gemini 4. "Getting that feedback from Googlers who are using the model every day, I think has been both helpful in terms of just better understanding the model and better understanding its strengths and weaknesses."
Doshi said Argon was helping to speed up robotics development inside Google, thanks to the model's improved multimodal and reasoning skills. Google said it trained a robot to play cornhole by having Argon watch a human player and calculate joint movements and spatial coordinates.
Google also said it has stepped up efforts to prevent the model from behaving in ways it shouldn't, employing a system that tracks the model's chain of thought to determine if it's behaving. "I think that's actually really important because then you can actually intervene. You can say, 'Okay, the model is actually veering off in a way that's concerning. Let's stop here,'" said Doshi.
"I think that gives us a much more robust set of protections in launching this model."
While Google employees are already testing better versions of Argon internally, Doshi said the model rolling out will be the one the company announced benchmarks for last month. However, Google is making clear that this is just the start of its next big push.
"People should expect to see continued improvements coming from us wholesale," Doshi said.
Have something to share? Contact this reporter via email at hlangley@businessinsider.com or Signal at 628-228-1836. Use a personal email address and a non-work device; here's our guide to sharing information securely.