Alphabet introduced Gemini 4 Argon as its most advanced AI model release, describing frontier performance across software engineering, enterprise knowledge and cybersecurity defense. The model is being used internally for memory efficiency at Google data centers, quantum computing optimization and other applications, with broader access planned through a phased rollout after trusted-user testing. Argon is initially priced at $2 per million input tokens and $10 per million output tokens. A Bloomberg report said some employees viewed Gemini 4 as weaker on practical tasks, including certain coding work, despite strong benchmarks; Google disputed that characterization, while a Google employee said there was "large consensus" internally that the model was at the frontier.