Gemini 4 Carbon expands access for complex coding and long-horizon agent workflows

Google expanded access to Gemini 4 Carbon on November 10, 2026, following its earlier limited rollout to cybersecurity defenders through Fairwind.

Model focus

Google positions Carbon for complex coding, debugging and open-ended research. It can spend more thinking tokens than Argon. Evaluate task success, elapsed time and total cost before switching workloads.

Multimodal context

The model card lists text, image, audio and video inputs with text output. The combined input-output context window is up to one million tokens; this is not a separate one-million-token output allowance.

Pricing and access

Period	Per million input tokens	Per million output tokens
Introductory offer	USD 2	USD 10
After the offer expires November 30, 2026	USD 4	USD 20

These announced rates apply to the publicly accessible Carbon model. Paid developers can use Gemini API entry points including AI Studio. Access is rolling out to Google AI Ultra subscribers across the Gemini app, Antigravity and Search AI Mode; availability depends on the account and product.

TboxIO perspective

Test real debugging, cross-file changes or analysis tasks, recording quality, verification effort and token usage. This report summarizes official materials; TboxIO has not independently tested the model.