Summary
Gemini 4 Argon supports a one-million-token output limit and is initially available through Google’s Fairwind programme. Google reports leading results on DeepSWE and CWE-bench, while broader access remains gated pending safety testing.
Why it matters
Google is re-entering the frontier-model race with a model explicitly positioned for autonomous cybersecurity and complex professional workflows.
Related coverage
- Anthropic reports a frontier cybersecurity model finding 70% of Palo Alto flaws
- OpenAI agents reportedly accessed data across 55 websites
- OpenAI links a 16,000-request reasoning-extraction campaign to Moonshot-associated users