Google’s Gemini 4 Argon arrives in Antigravity for AI work
Google says Gemini 4 Argon is now available to developers and enterprises through its API, and to AI Ultra subscribers in Google Antigravity. The company…
Google says its new Gemini 4 Argon model is available through Google Antigravity, bringing its latest AI system to developers, businesses and researchers working on complex tasks. The rollout also reaches subscribers of Google’s AI Ultra and Pro plans through the Gemini app and AI Mode in Search.
It is built for demanding work. Google says Argon handles tasks in areas including coding, finance, legal work, cybersecurity and science, and ranks first on the Vals Index, a benchmark covering several economic sectors.
Google describes Argon as a model for multi-step work rather than simple prompts. Its release gives developers and enterprises API access through AI Studio and the Gemini platform, while Google AI Ultra subscribers can use it in Antigravity.
Google cites coding gains from Antigravity
Google said its own engineering teams used Argon to tackle large codebases and optimize software. In Antigravity, the model helped teams increase agentic code submissions by more than 35% while reducing code rollbacks by nearly a third, according to the company.
The figures are Google’s account of internal use, not an independently reported industry-wide result. Still, they point to the practical stakes for organizations considering AI coding tools: faster code production matters less if changes create extra review work or need to be rolled back.
One example involved Chromium, the browser project. Google said Argon helped the Chrome team separate core webpage handling from a legacy folder of browser features, working across a codebase described as 40 million lines. The refactor spanned hundreds of files and, according to Google, was completed in weeks after months of planned work.
Short version: fewer crashes.
Google says the architectural changes are intended to let live webpages move between tabs, side panels and floating windows without reloading, while eliminating an entire category of browser crashes. The company did not give a release date for that functionality.
Model used across software and research
Google also described work on its open-source Agent Development Kit, a framework for enterprise agents. The team gave Argon its rulebook and a goal to reduce framework overhead without breaking existing applications. The model worked autonomously overnight and surfaced changes now live in the google/adk-python project.
Google reported that those changes cut in-memory session preparation time by 40% in long sessions and sped up cold imports by 21%. The company also said researchers used Argon in Antigravity to optimize xTB, an open-source quantum chemistry package used to predict molecular structures and screen drug and materials candidates. Their changes made it up to 2.3x faster at predicting molecular structures and 1.8x faster at conformer search.
For AI developers weighing access, Google announced an introductory price of $2 per million input tokens and $10 per million output tokens. Cached input tokens are priced at 95% off, according to the company.
Google’s claims extend to its Gemma models: it reported a 2.85x end-to-end inference speedup on an A100 via vLLM, with throughput rising from ~37,900 to over 108,000 tokens per second on Gemma 4 E2B. Google said the test kept 100% token parity and recorded zero quality drop.
Source: blog.google
