Skip to main content

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

Google DeepMind's Gemini 4 Argon: A Million-Token Monster

What if an AI could write a million words without stopping? That's the reality with Google DeepMind's latest model, Gemini 4 Argon, unveiled on September 30. Its output limit has skyrocketed from 64K to a staggering 1 million tokens—and to be clear, that's not the context window; it's how much the model can generate in one continuous stream. For tasks like deep codebase migrations, lengthy technical reports, or multi-step automation, this means no more chopping things into pieces. Most cutting-edge models today max out around 128,000 tokens, so Argon's leap is nothing short of revolutionary.

Image

Built for the Long Haul

Argon isn't your average chatbot. Google is rolling it out in phases: first to trusted cybersecurity pros, then developers, enterprises, and subscribers. It's designed for complex, long-term workflows in software engineering, law, finance, and cybersecurity defense. And the numbers speak for themselves: DeepSWE v1.1 hit a record 77.9% on long-term software engineering tasks; Vals Index (weighted for finance, programming, law, and taxation) ranks it first; Zapier's AutomationBench gave it 51.3%, also first; and long video understanding LVBench reached 91.7%, the highest yet. On the visual side, it handles professional charts, extracts details from long videos, and performs cross-document actions.

Security That Fights Back

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

Pricing and Protections

Pricing comes in two tiers: during the introductory period, it's $2 per million input tokens and $10 per million output tokens; later, that jumps to $4 and $20. Before wide release, Google is beefing up defenses against abuse, prompt injection, misaligned reasoning chains, and sandbox escapes. Thousands of Google employees are already using Argon to debug code, design algorithms, and migrate massive codebases. In one case, it improved quantum algorithms by 40% in minutes.

In short, Argon isn't here to chat—it's here to tackle the hardest challenges professionals face.

Key Points

  • 1 million token output limit—a massive jump from 64K, enabling uninterrupted long-form generation.
  • Phased rollout: first to cybersecurity experts, then developers, enterprises, and subscribers.
  • Top benchmark scores in software engineering, finance, law, and long video understanding.
  • Security-focused: autonomously finds and fixes vulnerabilities; already caught a critical hospital software flaw.
  • Pricing: introductory $2/$10 per million tokens, later $4/$20.
  • Advanced defenses against abuse, prompt injection, and misaligned reasoning.