The Pulse

Google makes Gemini 4 Carbon available to paying users

Google has opened Gemini 4 Carbon to paying developers and enterprises after an initial release to cybersecurity defenders. The model is also rolling out to Google AI Ultra subscribers, and its introductory API price is set to expire on November 30.

AI.info Team · · fact-checked · against 2 sources ↓ · Published 12 Nov 2026

Paying developers and enterprises can now access Gemini 4 Carbon through Google AI Studio and Gemini for business. Google said on November 10 that the model was also rolling out to Google AI Ultra subscribers in Antigravity, the Gemini app and AI Mode in Search. The broader release follows an initial rollout to a set of cybersecurity defenders through the Fairwind Program two weeks earlier.

Access and pricing

The introductory API price matches Gemini 4 Argon at $2 per million input tokens and $10 per million output tokens. After the introductory period expires on November 30, 2026, the price will rise to $4 per million input tokens and $20 per million output tokens. The price change applies to the Carbon model intended for public access.

In addition to AI Studio, developers can use the paid Gemini API tier in Android Studio. Enterprises can try it through the Gemini for business API or model selector, as well as in Google Antigravity or CodeMender.

Scientists and researchers can apply for a chance to use both Argon and Carbon through Antigravity’s highest usage tier free of charge for six months. Onboarding is rolling and subject to eligibility, capacity and selection. Cybersecurity defenders can apply to the Fairwind Program for rolling access to Carbon for cyber.

Model and agent tools

Carbon is based on Gemini 4 Argon, according to its model card. It accepts text, images, audio and video and produces text, with a combined input-and-output context window of up to 1 million tokens. The model is intended for tasks including multi-step coding and debugging and open-ended scientific research; it uses more tokens than Argon for extended reasoning and verification.

Google is also previewing /swarm in Antigravity, a feature that turns a prompt into a coordinated team of agents whose work can be tracked in real time. Developers can assign Carbon to architectural design, coordination and auditing while pairing it with faster, lighter models. The preview also uses parallel isolated workspaces and adversarial verifiers that test changes.

Benchmarks and limitations

Google said Carbon ranked first on CursorBench 4.0, Perplexity’s WANDR and Zapier’s AutomationBench, at more than 2X lower cost per task than the next-best external model. On a multimodal knowledge benchmark developed with Artificial Analysis, the company said Carbon ranked second at more than 6X lower average cost per task than the top-performing model. The benchmark tests multimodal performance in agentic workflows.

In an internal agentic honesty benchmark with 375 tasks, Carbon passed 92%, compared with 70% for Argon, according to Google. The company said Carbon declined to fill in missing experimental data 98% of the time and flagged methodological flaws 68% of the time.

Google reported that Carbon achieved a 0% prompt-injection attack success rate on the Gray Swan IPI benchmark. The company also said resistance to prompt injection remains an area for improvement. The model card notes that Carbon may hallucinate and can occasionally be slow or time out. Its knowledge cutoff is March 2026, though knowledge in some domains may be less current.