Google DeepMind made Gemini 4 Argon generally available on October 13, through the API in AI Studio and the Gemini platform for developers and enterprises, and for Google AI Ultra and Pro subscribers in the Gemini app, AI Mode in Search and the Antigravity development environment. The introductory API price is $2 per million input tokens and $10 per million output tokens, with cached input tokens discounted 95 percent. That undercuts the $4 and $20 pricing Meta set for Muse Wonder a day earlier, and it frames Argon as a work model: Google positions it for multi-step professional tasks in coding, finance, legal work, cybersecurity and science, and says it ranks first on the Vals Index, a benchmark covering economically significant work in those fields.
Google’s evidence for capability is mostly its own usage. Inside Google, the company says agentic code submissions rose more than 35 percent with rollbacks down nearly a third, and it cites Argon completing a Chromium refactor across hundreds of files, cutting session preparation time in its Agent Development Kit by 40 percent, and speeding inference on its open Gemma models by 2.85 times. In robotics, Google reports an 87 percent score on the ERQA embodied reasoning benchmark and a demonstration in which Apptronik’s Apollo humanoid learned a cornhole throw from video. These are company-reported results, not third-party evaluations, and the pattern matches every frontier launch this year: the vendor’s internal wins arrive first and independent replication later.
The commercial structure matters as much as the model. Argon is gated behind paid tiers for consumers, priced for enterprise API volume, and bundled into Antigravity, Google’s agentic coding environment. Google is also giving 10,000 scientists the top Antigravity tier free for six months, a land-grab for the research workflows where it claims early wins in quantum computing, fusion and rare disease work.
Coming one day after Meta’s frontier debut, Argon keeps the model race on a weekly cadence and shifts the competition toward price per completed task. The open question for the sector is the same one the IMF raised the same day: whether enterprise revenue grows fast enough to pay for the compute behind launches like this one.