Google launched Gemini 4 Argon for selected cybersecurity partners, priced at $2/million input tokens and $10/million output tokens.

Google’s DeepMind introduced Gemini 4 Argon, initially making it available to selected trusted cybersecurity defenders through the Fairwind program. The company described later access for developers, enterprises, consumers, paid API customers, and Google AI Ultra users as planned expansion following further safety work, not availability already delivered to all those groups.

Initial pricing is $2 per million input tokens and $10 per million output tokens, with cached input tokens priced 95% below the input-token price. The report says Google supports an output length of up to 1 million tokens, compared with 64,000 previously; this is explicitly an output-length claim, not a stated input-context limit.

Google reports 77.9% on the DeepSWE v1.1 software-engineering benchmark and 68% on the CWE-bench v1 vulnerability-remediation benchmark, describing the latter as tied for first place. It also claims leading results in financial research, legal research, and enterprise automation, but the supplied report does not provide numerical scores for those evaluations or independent validation.

Reported internal applications include code development, research, quantum-algorithm optimization, and analysis of data-center operating data that Google says freed more than 300 tebibytes (TiB) of memory. Google also described continuing work on misuse prevention, prompt-injection defenses, model-behavior monitoring, and security sandboxes, while a separate message says it participates in the US government’s voluntary model process.