Google has officially launched Gemini 4 Argon, its latest AI model designed to handle demanding tasks across software development, research, finance, law and cybersecurity. The release marks a fresh attempt by the company to strengthen its position in the increasingly competitive AI market, where OpenAI and Anthropic have made significant advances.
Unlike tools focused mainly on answering questions or generating content, Argon is built to work through complicated, multi-step assignments. Google said that the model can reason across longer workflows, follow instructions more reliably and produce more direct responses with less unnecessary repetition.
According to Google, Gemini 4 Argon performs strongly across several complex professional tasks, including software engineering, financial analysis and legal work. The company has also highlighted its ability to identify software vulnerabilities and help engineers investigate technical problems.
Google has already used the model internally for code optimisation, debugging and improvements to its infrastructure. One reported example involved reducing the computing resources needed to support YouTube's comments system.
These applications illustrate Google's broader ambition: developing AI that can contribute to technical work rather than simply provide suggestions. However, the company's performance claims are based on its own evaluations and should be considered alongside independent testing.
The model is rolling out through several Google products, but access depends on the subscription. Google AI Pro and AI Ultra subscribers can use Argon in the Gemini app and AI Mode in Search. Access through Google's Antigravity coding platform is currently more restricted, requiring an AI Ultra subscription.
Developers and businesses can also access the model through Google's paid API. This gives companies a way to integrate Argon's capabilities into their own applications and automated workflows. Google has not announced broad free access, leaving users on lower-tier plans with uncertainty over when they might be able to try the model.
Also Read: Meet Google’s New Gemini Coworker Built to Handle Workplace Tasks
Google has introduced an initial API price of USD 2 per million input tokens and USD 10 per million output tokens. From December 1, 2026, those rates are scheduled to increase to USD 4 and USD 20, respectively.
The launch follows a cautious initial rollout to selected cybersecurity specialists through Google's Fairwind programme. The company is expanding access while continuing to assess the risks associated with a powerful model that can perform advanced technical tasks.
For Google, Argon is an important step in its effort to compete at the frontier of AI development. Whether it can win over developers and businesses will depend on real-world performance, reliability, access and the cost of running demanding workloads.