Google DeepMind unveiled its newest frontier artificial intelligence model, Gemini 4 Argon, featuring an output capacity of 1 million tokens designed to execute long-horizon software engineering, enterprise research and cybersecurity operations.
The model increases maximum output generation from the previous limit of 64,000 tokens, enabling autonomous agents to sustain multi-step reasoning tasks in a single prompt execution. Introductory pricing for API access is set at $2 per 1 million input tokens and $10 per 1 million output tokens, with cached input tokens discounted by 95%.
Google has begun rolling out the model to select cyber defenders under its Fairwind Program before expanding availability to enterprise customers, software developers and Google AI Ultra subscribers.
In performance evaluations, Gemini 4 Argon recorded a score of 77.9% on the DeepSWE v1.1 benchmark for complex software engineering tasks and placed first on Zapier’s AutomationBench with a 51.3% score. On cybersecurity metrics, the model tied for first on CWE-bench v1 with a score of 68% and achieved 91.7% on the LVBench benchmark for long-form video interpretation.
Google reported using the system internally for large-scale codebase migrations, including converting C and C++ libraries to Rust across system components such as the Fuchsia Zircon kernel. Internal deployment of the model for memory optimization across company data centers identified savings between 500 tebibytes and 1 petabyte of memory.
“Built to sustain deep reasoning across complex, long-horizon workflows, Argon is fundamentally changing the way we work and build at Google,” Koray Kavukcuoglu, senior vice president at Google DeepMind and chief AI architect at Google, said in the announcement.
For defensive cybersecurity applications, Google is deploying Gemini 4 Argon without standard cyber guardrails to trusted security partners and internal teams, allowing the model to autonomously locate, test and patch software vulnerabilities. Cybersecurity firm Wiz utilized the model under its Scan for Good initiative to identify a critical data-exposure vulnerability in global hospital software.
To address safety risks associated with high-capacity models, Google stated it is conducting pre-release evaluations with the U.S. government. The company has integrated internal activation monitoring to spot potential misuse in chemical, biological, radiological and nuclear domains, along with isolated sandboxing environments to prevent autonomous execution drift.
This article is published on BitPinas: Google’s New Gemini 4 Argon Can Rewrite Entire Codebases and Patch Vulnerabilities in a Single Prompt
What else is happening in Crypto Philippines and beyond?
Be the first to comment