Google has launched Gemini 4 Argon, its latest and most advanced AI model, designed for deep reasoning on complex tasks. The model handles finance, coding, creative writing, and cybersecurity defense, with Google positioning it alongside frontier models like Open AI's GPT-6 Astra and Anthropic's Opus.
According to Artificial Analysis, Gemini 4 Argon matches GPT-6 Astra's Intelligence Index score at 60% of the cost per task. Currently priced at $2 per million input tokens and $10 per million output, Argon also scored one point ahead of GPT-6.1 Sol, with a 15% hallucination rate—the lowest among leading models.
Google is already using Argon for quantum computing research, codebase migrations, and memory optimization across its data centers, freeing up 300 Ti B of memory. The model features a 1 million token output limit and excels in visual understanding, including chart analysis and video detail identification.
Argon demonstrated autonomous vulnerability detection in healthcare software and tied for first place on the CWE-bench leaderboard with Grok 4.7 and GPT-6 Astra. Built with prompt injection resilience and misalignment mitigations, Argon is rolling out to Google's Fairwind Program members, with broader availability planned for developers, enterprises, and AI Ultra subscribers.
Source: Engadget · Summarized by HeadlinesBriefing