What Happened

On October 1, 2026, Google DeepMind launched Gemini 4 Argon, described as a frontier model for long-horizon coding, legal and finance work, and cyber defense. Access starts with trusted defenders in Google's Fairwind Program and U.S. government pre-release programs, then expands to paid API users and Google AI Ultra.

Initial pricing is $2 per million input tokens and $10 per million output tokens, rising later to $4 and $20. The model offers a 1M-token output ceiling and heavily discounted cached input.

Three Takeaways for Businesses

  • Budget for the price step-up. An introductory rate that doubles later changes your unit economics. Model your costs at the later price before you build on it.
  • Design for cached input. Heavily discounted cached input rewards workflows that reuse the same context, such as a fixed knowledge base or policy document.
  • Stay model-agnostic. Staged rollouts and changing prices are a reminder to keep your AI layer portable so you can switch providers without a rebuild.
The Bottom Line

A new frontier model is not automatically the right choice for your product. Test it against your real tasks, price it at the later rate, and keep an easy path back to alternatives.

What To Do This Week

List the AI workflows you run today, estimate monthly token volume, and compare cost at $2/$10 versus $4/$20. If the later price breaks your margin, plan an alternative before you commit.