Google announced Gemini 4 Argon on 13 October and described it as built for "deep work in complex, real-world tasks" in coding, finance, legal work, cybersecurity and science.
Who can use it
- Developers and enterprises, through the API in AI Studio and Google's Gemini platform.
- Google AI Ultra subscribers, in Google Antigravity.
- Google AI Ultra and Pro subscribers, in the Gemini app and AI Mode in Search.
- 10,000 scientists can apply for the highest Antigravity tier free for six months, subject to eligibility and selection.
Cached input tokens are 95% cheaper, and the API has a 1M unified input-output token limit. Access without the usual cyber guardrails goes to vetted defenders through what Google calls the Fairwind Program, on a rolling basis.
What Google claims
- First place on the Vals Index, which covers finance, coding, legal and tax work.
- 87% on the ERQA embodied reasoning benchmark, which Google says beats every other frontier model.
- A 0.7% attack success rate on the Gray Swan indirect prompt injection benchmark, across 15 adaptive attempts.




