Skip to main content

Gemini 4 Argon: Google's New AI Writes a Million Tokens in One Go

Google DeepMind quietly rolled out its latest powerhouse, Gemini 4 Argon, on September 30. The headline-grabbing feature? An output limit that jumps from the previous 64K to a whopping 1 million tokens. That's not about how much it can read—it's about how much it can write in one continuous stream. Imagine generating a full-length technical report or migrating an entire codebase without breaking it into chunks. Most top models today cap out around 128,000 tokens, so Argon's leap is like going from a sprint to a marathon.

Built for the Long Haul

Argon isn't your average chatbot. Google is positioning it for "complex, long-term workflows" in fields like software engineering, law, finance, and cybersecurity. The rollout is phased: first trusted security pros, then developers, enterprises, and subscribers. The benchmarks back up the hype. On DeepSWE v1.1, which tests long-term software engineering, it hit a record 77.9%. It also topped the Vals Index (a weighted mix of finance, programming, law, and tax) and scored 51.3% on Zapier's AutomationBench. For long video understanding, it reached 91.7% on LVBench—the highest yet. And it can handle professional charts, extract details from lengthy videos, and perform cross-document actions.

Security That Fights Back

Cybersecurity is the beating heart of this release. Argon was trained to autonomously find, verify, and fix critical vulnerabilities. For trusted defenders and Google's internal teams, there's even a version with no network barriers. In one case, security firm Wiz used it through the Scan for Good project to uncover a serious flaw in a global hospital's medical software—something previous cutting-edge models missed. On CWE-bench v1, which tests vulnerability repair, it ranked first with 68%.

Image

Pricing and Protections

For early adopters, pricing is set at $2 per million input tokens and $10 per million output tokens. After the introductory period, those rates double to $4 and $20. Before opening the floodgates, Google is beefing up defenses: preventing abuse, blocking prompt injection, monitoring misaligned reasoning chains, and strengthening sandboxes. Thousands of Google employees are already putting Argon through its paces—debugging code, designing algorithms, and handling massive code migrations. In one instance, it improved quantum algorithms by 40% in minutes. Clearly, Argon isn't here for small talk. It's built to tackle the toughest challenges professionals face.

Key Points

  • Million-token output: Argon can generate up to 1 million tokens in a single output, far beyond the typical 128K limit.
  • Targeted at professionals: Designed for complex, long-term tasks in software, law, finance, and cybersecurity.
  • Top-tier benchmarks: Leads in long-term software engineering (77.9%), Vals Index, AutomationBench (51.3%), and LVBench (91.7%).
  • Security focus: Autonomously finds and fixes critical vulnerabilities; already used to detect a hospital software flaw.
  • Tiered pricing: $2/$10 per million input/output tokens initially, doubling later.