Gemini 4 Argon is finally here, and it's not just for coders and developers

A fortnight after teasing its latest and potentially greatest AI model, Google is finally bringing Gemini 4 Argon to everyday users. When first revealed, the AI model seemed to boast highly competitive benchmark scores across a wide range of agentic tasks. Now that the model is finally out, Google is giving users a chance to put those claims to the test.

Argon 4 offers major improvements in coding (compared to its predecessors), but you don’t necessarily need to be a coding enthusiast or a developer to benefit from the latest AI model. Argon is better at regular tasks too.

According to Google, the new model is better at understanding complex questions, real-world tasks, and work in domains like finance, legal, cybersecurity, and science. It also boasts stronger writing capabilities that allow it to communicate clearly with less hedging, low repetition, and the sense to avoid clichés. It has better understanding of your preferred writing style, so you wouldn’t necessarily need to repeatedly remind it to use a professional tone for emails or a casual one for social media posts.

Google says Argon follows style-based instructions better than any of the previous Gemini models. This should prevent the need to bring Gemini back to the topic during extended sessions where it strays off conversation.

Apart from everyday use, Google claims that Argon ranked #1 on the Vals Index, which measures performance across economic sectors like finance, coding, legal, and tax work. The folks over at Google have already been using the model to solve engineering problems, optimize business workflows, find and patch cybersecurity vulnerabilities, and to potentially find breakthroughs in quantum computing, nuclear fusion, and life sciences.

Using Argon in Antigravity has reportedly helped increase agentic code submissions by more than 35%, all while reducing code rollbacks by nearly a third. Essentially, engineers at Google are now relying on AI for coding a lot more than ever before.

Argon is smarter, but also safer

Given its advanced capabilities, the tech giant is also emphasizing safety. According to Google, Argon is adept at refusing potentially dangerous requests, including prompts to generate malicious code or instructions involving dangerous biological materials. Elsewhere, the model is better at respecting system boundaries. For example, if it encounters a roadblock, it is designed to seek human approval rather than find an unsafe workaround. These are Google’s own testing results and claims. Whether they stand true in real-world usage remains to be seen.

Gemini 4 Argon is rolling out now. If you’re an everyday user, you can try out the new model in the Gemini app or AI Mode in search, provided that you’re a Google AI Ultra or Pro subscriber. Developers can access the model via the Gemini API paid tier in AI Studio, Android Studio, and Google Antigravity. Google said it’s looking forward to introducing the rest of the Gemini 4 family.