Google Opens Access to Gemini 4 Carbon
Google has made Gemini 4 Carbon available for long-horizon coding agents and hypothesis testing, with paid API access now and Ultra-only availability in Gemini chat and Antigravity.
Google Opens Access to Gemini 4 Carbon
Gemini 4 Carbon is built on Argon for tasks where an agent needs to dig through code for a long time or test hypotheses. It consumes more tokens for reasoning than Argon. In a Google demo, the model assembled an interactive 3D simulator that synthesizes the songs of six bird species on the fly; users can even open a beak and listen to how the sound changes.
On the new MKA benchmark, agents receive a set of blueprints and other files and must assemble a report, table, or presentation. Under complex criteria, Carbon High scores 54%, while Fable 5.2 Max scores 57%.
The model is already available in the paid API, and in Gemini chat and Antigravity it is offered to Ultra subscribers. Until the end of November, API pricing is $2/$10 per million input/output tokens; from December 1 it rises to $4/$20.
Why it matters
AnalysisThis matters because Google is expanding its agentic AI stack for software work, and the upcoming API price change will directly affect the economics of code-heavy and reasoning-heavy workflows.
Discuss in community
Share your questions and insights with developers
Related in Ecosystem
Practical guides, core terminology, and agent skills
Google releases Gemini 4 Carbon
Google announced the Gemini 4 Carbon AI model and made it available to a wider audience, effectively limited to AI Ultra subscribers paying $250.
Google releases Gemini Decision 1, an AI model for decision-making
Google has introduced Gemini Decision 1, a decision-focused AI model available through the Gemini API with multimodal input support and a 256K-token context window.
Google AI Studio: Free Gemini API Keys
Official Google sandbox: get free Gemini 2.0/1.5 API keys without a credit card, setup system prompts, and test the 2M token context window. Limits & guide 2026.
Gemini Flash & Pro (Google Gemini)
A family of multimodal models from Google DeepMind that combines a record context window (up to 2 million tokens), extreme generation speed (over 150 tokens/sec), and native perception of video and audio.