Google releases Gemini 4 Argon, says it matches rivals on coding and cybersecurity benchmarks

Google released its new-generation flagship model Gemini 4 Argon, saying it reaches SOTA on benchmarks in coding, enterprise office work, science and mathematics, and cybersecurity, and can match Anthropic and OpenAI competitors. The model's context window has been expanded 16 times, and Google has already used it internally for quantum computing research, data center memory optimization, and automated programming. Argon is first opened to some enterprises and governments through the Fairwind program to identify and fix cybersecurity vulnerabilities in advance, and Google will fine-tune the guardrails accordingly; the company says the model is more resistant to prompt injection and has added protections to prevent agents from escaping the test environment. After public release, base pricing is $2 per million input tokens and $10 per million output tokens, rising to $4 and $20 after the promotional period. Previously, Google abandoned 3.5 Pro because training did not meet expectations, and instead launched Argon directly.