Google's AI Leap: Gemini 4 Argon Outperforms Rivals

Google has officially rolled out its cutting-edge artificial intelligence model, Gemini 4 Argon, making it accessible to consumers and enterprises worldwide. This advanced AI is now available through the Gemini API, the Gemini app, and an AI Mode in Search, following a period of exclusive testing.

Google positions Gemini 4 Argon as a significant response to competitors such as OpenAI's GPT-6 Astra and Anthropic's Claude models. Benchmark results indicate Gemini 4 Argon leads in various domains, including coding, finance, legal work, and cybersecurity, notably achieving the top rank on the Vals Index.

The new model shows strong performance in complex, multi-step tasks crucial for enterprise workflows. It particularly excels in agentic coding, science, math, and long-context understanding, with a 99.7% score on the GraphWalks benchmark.

Specific use cases highlight Argon's impact, such as boosting agentic code submissions by over 35% in Google's Antigravity platform. Internally, the Chrome team utilized Argon to accelerate a complex code separation task, completing months of work in weeks.

In cybersecurity, Gemini 4 Argon ties for the top score on the CWE-bench v1, demonstrating its ability to identify and remediate vulnerabilities. It also significantly sped up scientific computations, enhancing the xTB quantum chemistry package by 2.3x.

Google emphasizes Argon's improvements in spatial reasoning, mathematical precision, and visual coding. The model's public access is priced at an introductory rate of $2 per million input tokens and $10 per million output tokens.

This launch follows Google CEO Sundar Pichai's participation in a voluntary White House Accord on Super Intelligence, alongside other AI leaders who pledged enhanced controls and monitoring mechanisms.