Gemini 4 Argon: 33% cheaper than GPT-6.1 Sol, up to 1M tokens per response

Google's Gemini 4 Argon generates up to 1 million tokens in a single response and cuts operational costs by 33% versus GPT-6.1 Sol and 70% versus Claude Opus 5.5. The model targets enterprise knowledge work, large-scale codebase migrations and vulnerability detection.

• Up to 1M tokens per response via dynamic compute allocation at inference
• Operational costs 33% below GPT-6.1 Sol and 70% below Claude Opus 5.5
• Tops Waltz, Zapier Automation Bench and CWE bench leaderboards
• Use cases: C/C++ to Rust migrations, data analysis, cybersecurity