Gemini 4 Argon Lands at $2 and $10 per Million Tokens, Matching GPT-6 Sol

Google priced Gemini 4 Argon at $2 per million input tokens and $10 per million output tokens — the same rate OpenAI lists for GPT-6 Sol, and half the $4 and $20 Anthropic charges for Claude Opus 5.5. Google DeepMind announced the model on September 30 with introductory pricing and a 95 percent discount on cached inputs.

The claims are aimed at professional work. Google says Argon sets a new state of the art on DeepSWE v1.1 at 77.9 percent, ties for first on CWE-bench v1 at 68 percent for vulnerability remediation, ranks first on AutomationBench at 51.3 percent and leads the Vals Index of economic impact across finance, legal and tax work. It also offers what Google calls an industry-leading 1 million token output limit. These are Google’s own figures; OpenAI’s GPT-6 Sol launch reported 68.8 percent on the same DeepSWE v1.1 benchmark, and Anthropic’s Opus 5.5 launch is priced at $4 and $20 per million tokens.

Google says it used Argon internally for quantum optimization work, memory efficiency improvements that freed more than 300 terabytes, and codebase migrations totaling hundreds of thousands of lines. Security firm Wiz used it to find critical healthcare vulnerabilities that earlier models missed.

Access is staged. The rollout begins through Google’s Fairwind Program for cyber defenders, with safety testing before broader developer and consumer availability. The price is a reference point today, not a purchase option for most teams.

The practical move: when broader access opens, run the same task set through all three models. Three vendors now sit within a factor of two on list price, so cost per finished task on your own work, not the benchmark chart, decides which one ships.