Industry Odisha Bureau, Oct 1: Google has launched the all new Gemini 4 Argon, its latest AI frontier model. The company claims major improvements in complex reasoning, software engineering, professional workflows and cybersecurity.
Google CEO Sundar Pichai took to his official handle on X to announce the model. He said that Argon delivers frontier performance in complex workflows, cyber defence and software engineering. He also said that Google teams are already using the model for tasks including coding and quantum computing.
Notably, Gemini 4 Argon comes with a 1 million-token output limit. Google says that it allows the model to handle longer and more complex tasks by generating hundreds of thousands of tokens in a single trajectory.
Google’s benchmark results show Argon performing strongly against competing AI models in several areas. On the Vals knowledge-work benchmark, it scored 68.9 percent, compared with 63.1 percent for GPT-6 Astra and 65.8 percent for Claude Fable 5.1. On the Vibe code Bench, Argon scored 91.9 percent. This was again ahead of Fable 5.1 at 90.3 percent and GPT-6 Astra at 89.6 percent. Google noted that Argon trails some rivals on certain tests, including Terminal bench 4.0.
Cybersecurity is another major focus. Google says Argon can autonomously identify, validate and patch critical vulnerabilities. Its cybersecurity unit Wiz has already tested the model and reported that it identified a critical vulnerability in healthcare software that earlier frontier models had missed.
For now, Argon is being made available to the US government and selected cybersecurity defenders through Google’s Fairwind programme. Broader access will initially target paid API customers and Google AI Ultra subscribers.
Google has set an introductory price of $2 per million input tokens and $10 per million output tokens.
