Franklin AI Analysis

Gemini 4 Carbon opens to paying developers

Key Takeaways

  • Google opens Gemini 4 Carbon to paying developers and enterprises and starts an Ultra rollout.
  • The company lists the same introductory API rates as Argon.
  • Google expanded access to Gemini 4 Carbon after a limited release.
  • It lists an introductory API price of $2 per million input tokens and $10 per million output tokens, with a separate rollout for Google AI Ultra subscribers.

Google announced broader availability for Gemini 4 Carbon on November 10 in a company post. The API is available to paying developers and enterprises, and Google is rolling the model out to Google AI Ultra subscribers. The company positions Carbon for long tasks that require extensive reasoning and coordination.

Who can access Carbon

Google names Google AI Studio and Gemini for business for API access. It names Google Antigravity, the Gemini app and AI Mode in Search for the Ultra rollout. The company says an earlier release was limited to cybersecurity defenders in its Fairwind Program. Access depends on the product and account involved, so a team should check the interface it actually uses.

Pricing and project scope

Google lists $2 per million input tokens and $10 per million output tokens as introductory API rates, the same as Gemini 4 Argon. The company also previewed Antigravity’s swarm feature, which lets developers assign work to coordinated agents. It describes using Carbon for roles such as architectural design, coordination and auditing while assigning other work to different models.

What to test before switching

The announcement describes examples in which Carbon handles dense documents, complex engineering and multi-step projects. Those examples illustrate what Google tested under particular conditions. Token rates are only part of a workload’s cost because longer runs can consume more tokens. Teams choosing a model can compare the completed outcome, time and total cost on the same tasks, keeping the underlying prompts and evaluation criteria consistent across trials.

Our read

Franklin AI Take

Carbon and Argon start at the same published token rates. That makes a side-by-side project test useful, particularly when several agents can each consume tokens. Pick a bounded task with a verifiable output, then record time, total token use and successful completion. Check availability in the specific Google product your team plans to use.