Google has begun a global rollout of Gemini 4 Argon, its latest frontier model, with access split across its developer tools and consumer products. Google's announcement says developers and enterprises can use Argon through the paid Gemini API tier in AI Studio and the Gemini platform. Google AI Ultra subscribers get access in Antigravity, while Pro and Ultra subscribers can use it in the Gemini app and AI Mode in Search.
The company is positioning Argon for long, multi-step work in coding, finance, legal, cybersecurity and science. Its most conspicuous technical change is a unified 1 million-token limit shared across input and output, which Google says is designed for extended scientific, coding and agentic work. Google directs developers and enterprise customers to a new continuation protocol for setting up longer-running work.
Price and product access
Google set an introductory API price of $2 per million input tokens and $10 per million output tokens, with cached input discounted by 95%. The company says Argon ranks first on the Vals Index, a benchmark covering economic work such as coding, finance, legal and tax tasks. Those results come from Google's launch materials and should be read as vendor claims rather than independent validation.
The Gemini app announcement says Argon answers more directly, shares progress during longer tasks and improves coding. Google is also adding Antigravity's agentic coding interface to the Gemini app on macOS and offering a task mode that can keep working across connected services. The app supports more than 40 integrations, including Instacart, OpenTable, Adobe and Shopify, according to Google.
Early testing finds strengths and rough edges
In an early hands-on assessment, Every CEO Dan Shipper said six people spent a week testing Argon across coding, writing, design and knowledge work. Shipper called it his new favorite writing partner and praised its ability to organize essays and explain texts. His coding results were less consistent: Argon performed well on a new app but scored poorly on a large refactoring test, and in 17 of 38 benchmark runs it appeared to skip a meaningful code-review step.
Shipper also described Antigravity as behind competing coding agents in areas such as cloud environments, SSH access, browser integration and permission handling. His conclusion was narrower than Google's broad launch pitch: Argon looked useful for writing, prototypes and selected design work, but he preferred using the API or another coding interface for some tasks.
Comments