Google’s Gemini 4 Argon Bets Premium Access Can Put It Back at AI’s Frontier

Google has released Gemini 4 Argon to paying users and developers, touting gains in reasoning, coding and security. The launch is a bid to regain ground in an AI race that has grown far more crowded.
Google’s Gemini 4 Argon Bets Premium Access Can Put It Back at AI’s Frontier

Google’s Gemini 4 Argon Bets Premium Access Can Put It Back at AI’s Frontier
Google had already previewed Argon to a small group of “trusted cyber defenders,” using the limited release to probe for cyber risks before a wider debut. The timing matters: after Gemini 3’s late-2025 release, Google had watched newer models from OpenAI and Anthropic dominate the frontier-model conversation.

On Tuesday, the company opened Gemini 4 Argon globally to developers and enterprises through the paid Gemini API and platform, while Pro and Ultra subscribers gained access in the Gemini app and Search; Antigravity access is reserved for Ultra users. That means the model’s public arrival is still a premium one, with Gemini plans starting at $19.99 a month for app access.

Google’s case for a comeback rests on Argon’s claimed strength in coding, cybersecurity and professional knowledge work. Tulsee Doshi, Gemini product lead at Google DeepMind, called it a “powerhouse” and said it would lead a broader family of models. Internally, Google says Argon has raised engineers’ production of AI-generated code by 35%, while helping work on products including Chrome and robotics.

Pichai’s demonstrations stress the same pitch: stronger coding and multistep tool use, including an Antigravity project that turned a raw Pixel CAD file into an animated, staged presentation. He also highlighted a test in which the model processed more than 80,000 NASA Mars weather and Ingenuity helicopter telemetry logs into a presentation on atmospheric flight mechanics.

Google says safeguards were part of the path to release, including monitoring a model’s chain of thought for concerning behavior so operators can intervene. Its pricing starts at $2 per million input tokens and $10 per million output tokens—roughly one-fifth of cited rival rates, according to one account—before introductory prices double after November. Whether those claims translate into a durable lead will be tested against OpenAI, Anthropic and Meta—not by Google’s benchmarks alone.

https://foxvector.com

Write a comment