Google releases Gemini 4 Argon AI model globally

Google DeepMind has launched Gemini 4 Argon, the first model in the Gemini 4 generation, making it available globally via the Gemini API, Google AI Studio, the Gemini app, and Google Search’s AI mode. The model is specifically designed for complex, multi-step workflows in specialized fields such as programming, finance, law, cybersecurity, and science.

In technical benchmarks, Gemini 4 Argon achieved a 97.5% score on the ARC-AGI-3 general problem-solving benchmark using a provider adapter harness. It also ranked first on the Vals Index, which measures performance in critical economic sectors like coding and legal tasks. To support research, Google announced a program providing 10,000 scientists and researchers with six months of free top-tier access.

For developers and businesses, the model is being offered at an introductory price of $2 per million input tokens and $10 per million output tokens through November 30, 2026. After this period, prices are expected to double. The model demonstrates high efficiency in coding tasks, with internal Google teams reporting significant improvements in code submission volume and architectural refactoring speeds.