Ethical Hacking News
Google has unveiled its latest artificial intelligence (AI) model, Gemini 4 Argon, designed to support long-horizon software engineering and cybersecurity. The model is priced to start at $2 per million input tokens and $10 per million output tokens, with a focus on cybersecurity. Argon has already shown impressive results in various internal tests, including optimizing resource-heavy processes and finding memory optimizations. The model is now being used to rewrite C and C++ code in Rust and is being made available to trusted security teams and Google's own engineers. While Argon has the potential to revolutionize cybersecurity, Google is also taking steps to prevent misuse and ensure alignment with user intentions.
Google has unveiled its latest AI model, Gemini 4 Argon, which is being tested on its own infrastructure before its public release.Argon's output token limit has been increased to 1M tokens, up from 64,000 tokens on the previous generation.Argon is priced at $2 per million input tokens and $10 per million output tokens, with discounted prices for cached inputs.Argon has already shown impressive results in internal testing, including optimizing quantum computing processes and freeing up memory in Google's data centers.Argon is being used to rewrite C and C++ code in Rust and has improved the performance of open-source projects.Argon has achieved top rankings in various benchmarks, including DeepSWE v1.1, Vals Index, and Zapier's AutomationBench.Google is focusing Argon on cybersecurity and making it available to trusted security teams and its own engineers.Google is deploying measures to prevent misuse of Argon, including blocking misuse and deploying misalignment mitigations.
Google has unveiled its latest artificial intelligence (AI) model, Gemini 4 Argon, which is being tested on the company's own infrastructure as a first step towards its release to the public. The model is designed to support Gemini 4 Argon's capabilities across longer, more complex use cases, and its output token limit has been increased to an industry-leading 1M tokens, up from 64,000 tokens on the previous generation.
According to Google, Argon is priced to start at $2 per million input tokens and $10 per million output tokens, with cached inputs at a 95% discount. Once the introductory window closes, the price will jump to $4 and $20, respectively. This new pricing structure is intended to make the model more accessible to a wider range of users while still maintaining its value proposition.
Google engineers have already begun using Argon internally, and the results have been impressive. In one instance, Argon helped quantum computing researchers optimize a resource-heavy process, improving the published baseline by 40% in just a few minutes. In another instance, Argon agents analyzed data from Google's data centers and found memory optimizations that freed more than 300 TiB after deployment. Google estimates that there could be another 500 TiB to 1 PiB of savings.
Argon is also being used to rewrite C and C++ code in Rust, including Google's Fuchsia Zircon kernel, which has more than 800,000 lines of code. For the open-source libgav1 video decoder, Argon replaced 32,000 lines of SIMD code through repeated testing and compiler analysis, resulting in a version that runs 2.7 times faster than the previous Rust version, produces the same video output, and remains memory-safe.
In terms of benchmarks, Argon tops DeepSWE v1.1 at 77.9% for long-horizon software engineering, leads the Vals Index across finance, legal, and tax work, and ranks first on Zapier's AutomationBench at 51.3%. It also posts a state-of-the-art 91.7% on LVBench, a benchmark for understanding long video.
Google is focusing Argon on cybersecurity, making it available to trusted security teams and its own engineers, who can use it without cyber safety restrictions. This allows Argon to actively look for vulnerabilities that defenders can then fix. Wiz, a cybersecurity firm, is already using Argon through its Scan for Good program, which looks for serious vulnerabilities in public systems and fixes them for free. Argon found a critical flaw in healthcare software used by hospitals worldwide that could expose personal data, and earlier AI models had examined the same issue but failed to find it.
To prevent misuse of Argon, Google is hardening defenses on four separate fronts: blocking misuse for cyber or CBRN attacks while still allowing legitimate dual-use research, improving resistance to indirect prompt injection, and sealing off the sandboxed environments used for high-risk training and testing before anything risky happens inside them.
In order to prevent Argon from stepping out of bounds to try to accomplish a task in a way that goes beyond the user's intentions, Google is deploying misalignment mitigations that monitor Argon's chain-of-thought and actions and stop execution when necessary. Google strongly encourages the rest of the industry to preserve reasoning transparency in these pivotal moments of increased capabilities while navigating alignment risks, so that model thoughts remain helpful in identifying and diagnosing misalignment.
Related Information:
https://www.ethicalhackingnews.com/articles/Googles-Gemini-4-Argon-A-Revolutionary-AI-Model-for-Cybersecurity-and-Beyond-ehn.shtml
https://securityaffairs.com/200187/uncategorized/inside-gemini-4-argon-the-model-google-is-testing-on-its-own-infrastructure-first.html
Published: Thu Oct 1 09:45:52 2026 by llama3.2 3B Q4_K_M
