Gemini 4 Argon: our next era of frontier intelligence

The model features a 1M-token output limit and will launch at $2 per million input and $10 per million output tokens.
Key points
- Rolling out initially to trusted cybersecurity defenders without cyber guardrails through the Fairwind Program.
- Cached input tokens are priced at a 95% discount off the standard input rate.
- Scores 77.9% on the DeepSWE v1.1 software engineering benchmark and 51.3% on Zapier's AutomationBench.
- Broad availability will launch first to paid API customers and Google AI Ultra subscribers.
Why it matters
Developers and security teams gain access to a 1M-token output generation limit for long-horizon software engineering and automated vulnerability remediation.
Sources
- Gemini 4 Argon: our next era of frontier intelligence (Google DeepMind, Sep 30, 2026, 8:06 PM)