Artificial Analysis says Gemini 4 Argon (high) matches GPT-6 Astra (max) on its Intelligence Index and has a 15% hallucination rate, compared with 51% for Astra (Artificial Analysis)

Alphabet, Google's parent company, has introduced Gemini 4 Argon, a cutting-edge AI model designed to tackle a range of tasks including coding, research, and writing. However, Google asserts that Argon excels in cybersecurity, with the model reportedly capable of autonomously identifying, validating, and patching critical software vulnerabilities.

This specialized training is part of Google's Fairwind Program, a security initiative focused on cyber defense. The company has already seen internal adoption, with employees employing Argon for tasks such as debugging and managing codebases. Furthermore, Argon's ability to analyze visual data, whether it's long videos or complex charts, is another area where it excels.

Google emphasizes that Argon is designed to perform deep reasoning across intricate, long-term workflows, fundamentally transforming the way Google operates and builds. This move comes at a time when leading AI labs are vying to release increasingly powerful models, in an effort to surpass one another, despite warnings from the very same companies about potential AI risks.

Recently, OpenAI unveiled Astra, which it branded as its most advanced model yet. Meanwhile, Anthropic launched Fable, also boasting similar claims of superiority. In its blog post, Google claims that Argon outperforms OpenAI's GPT-6 Astra and Anthropic's Fable and Opus models across various AI benchmarks. The tech giant cites Vals, a popular AI benchmarking startup, as proof that Argon currently holds the top spot on Google's AI model index.

In recent years, Google has turned the tide in the "AI race," achieving a significant milestone with over a billion users per month for its Gemini app. This metric places Google on par with OpenAI, which recently announced that ChatGPT had also reached a billion monthly users.