Google adds Gemini 4 Carbon model to Gemini app for AI Ultra users

Google has rolled out its newest Gemini 4 Carbon model to the Gemini app for AI Ultra subscribers. The update lets users access Carbon in the app’s AI Mode and Antigravity features, while the existing Argon model remains available for AI Pro and higher tiers.

Google says Argon is "the ideal choice for the majority of software engineering and knowledge work," whereas Carbon is intended for "tasks that test the limits of a model’s intelligence," such as complex, multi-step coding, debugging, and open-ended scientific research. In the app’s model picker, Argon is labeled for "Code & analysis" and Carbon for "Most complex work," and the previous 3.1 Pro option has been removed.

The company also highlighted Antigravity’s new "swarm" capability, which can turn a single prompt into a coordinated team of agents that users can monitor in real time. Developers can assign Carbon to roles requiring deep reasoning—like architectural design, coordination, and auditing—while pairing it with faster, lighter models for parallel tasks.

For legal, finance, and enterprise applications, Google claims Carbon delivers higher-precision knowledge work, artifact generation, and data analysis, following strict schemas to produce polished spreadsheets, presentations, and reports. To evaluate performance on dense multimodal inputs, Google partnered with Artificial Analysis to create the Multimodal Knowledge Agents (MKA) benchmark. Carbon ranked first, using three times fewer tokens than the next best model across 200 tasks.

The introductory pricing for Carbon is $2 per million input tokens and $10 per million output tokens through November 30, after which the rates will double to $4 per million input tokens and $20 per million output tokens.