Google has officially launched Gemini 4 Argon.

Google has officially launched Gemini 4 Argon.
Summary
Google has launched the Gemini 4 Argon model, aiming to compete in AI.
Argon outperforms rivals in coding, cybersecurity, science, finance, and law benchmarks.
The model is initially available to Pro and Ultra subscribers at competitive pricing.

Share

Bookmark

Newsletter

A London-based journalist from The Verge specializing in AI, who also serves as a Senior Tarbell Fellow, has taken a significant step in the tech world. Previously, this reporter focused on health, science, and technology for Forbes, and their insights will now become a regular feature in your daily email roundup and on your homepage.

After a prolonged period of observing the evolving AI landscape from a distance, Google has unveiled its long-anticipated Gemini 4 Argon model, asserting that it stands shoulder to shoulder with leading competitors in the field.

According to Google, Argon showcases “frontier reasoning and enhanced performance across intricate workflows,” effectively reducing issues such as hedging, redundancy, and the clichés that often hinder AI chatbots. The tech giant claims that Argon surpasses competing models, including GPT-6 Astra, Claude Fable 5.1, and Claude Opus 5.5, across various benchmarks encompassing coding, cybersecurity, science, finance, and legal matters.

Google is already integrating Argon for internal applications, leveraging its capabilities to identify and resolve cybersecurity weaknesses, significantly reduce the computational resources required for managing YouTube’s comment section, and refine the drone physics simulator used by its delivery service, Wing.

The model is being rolled out today to Pro and Ultra subscribers of Google’s Gemini app and AI mode in Search, as well as to Ultra subscribers utilizing the Antigravity development platform. However, the timeline for availability to Plus subscribers or free users remains uncertain.

In terms of pricing, Google has positioned Argon competitively in the market. The initial pricing is set at $2 for every million input tokens and $10 for every million output tokens, approximately 20% of the cost associated with models like OpenAI’s GPT-6 Astra and Anthropic’s Fable 5. Notably, even when prices are expected to escalate after November’s introductory phase, Argon will still offer a tremendous cost advantage over its top-tier competitors on a per-token basis, though actual expenses may vary due to different token usage for comparable tasks.

This launch clears up some uncertainties that emerged when Google initially showcased Argon a few weeks back. Like other recent advancements in frontier AI models, the public release follows a limited rollout to a select group of “trusted cyber defenders” to assess potential risks.

Argon marks Google’s first significant frontier model release since Gemini 3 debuted in late 2025. During the time elapsed, outstanding models from Anthropic and OpenAI had enabled them to leave Google trailing. Internal upheavals within the company raised concerns regarding its capability—or motivation—to regain its footing in the competitive AI landscape.

However, by introducing Argon, Google demonstrates that it should not be overlooked in the AI race. Arriving just one day after Meta’s announcement of its own frontier model, Muse Wonder Full, Argon serves as a potent reminder of the rapidly shifting dynamics within the AI sector, underlining that the competition is robust and far from being a simple two-player contest.

Loading comments...