Google DeepMind and Google AI announced the global availability of Gemini 4 Argon, which the company describes as its latest frontier model. According to the announcements, Argon is accessible to developers and enterprises through the API in AI Studio and the Gemini platform, to Google AI Ultra subscribers in Google Antigravity, and to Google AI Ultra and Pro subscribers in the Gemini app and AI Mode in Search. Google states the model targets 'deep work in complex, real-world tasks' across coding, finance, legal, cybersecurity and science, and says it ranks #1 on the Vals Index, a benchmark the company says measures performance across finance, coding, legal and tax work. The company lists an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens at 95 percent off. Google also reports internal results, stating that software engineers using Argon in Antigravity increased 'agentic code submissions by more than 35%' while reducing code rollbacks by nearly a third. The announcement includes demonstrations such as building a 3D pagoda model and a miniature racing game via multi-agent loops. All figures and claims are drawn from Google's own announcements; independent verification is not provided in the source text.
What it means for you
A new, more capable model exists and is priced competitively per token. For most organizations, this changes nothing you need to act on today — it's a faster engine, not a new road. The benchmark wins and the '35% more code submissions' figure come from Google measuring its own teams, so treat them as marketing until you see results in your own work.
Try this
If your team already uses Gemini or its API, pick one real task you run regularly and compare the new model's output against what you get now — same input, side by side. Don't switch anything based on a vendor benchmark; judge it on your own work.
Who should care
Teams already building on Google's AI Studio or Gemini API, and whoever owns developer tooling if your engineers use Antigravity. Anyone paying for Gemini Ultra or Pro will see it in the app.
Skip this if
You don't currently use Google's AI tools, or you have no process yet for where AI fits in your work. A better model on top of an unmapped process just automates confusion — sort the process first.
Sources: Google AI, Google DeepMind — read the original