Gemini 4 Carbon (high) - Intelligence, Performance & Price Analysis

Why it matters

Artificial Analysis has published a detailed evaluation of Gemini 4 Carbon (High), a proprietary reasoning model released by Google in late October. The model scores well above the median on the Artificial Analysis Intelligence Index when measured against comparable models in its price tier, indicating strong performance in knowledge work, coding, science, and general knowledge tasks. However, the assessment highlights notable trade-offs. Its pricing sits at the higher end relative to similar models, and it is notably verbose, generating far more output tokens during evaluation than the median for its class. Speed is a mixed picture: output generation is slightly faster than average, but the time to first token is at the higher end, which affects responsiveness. The model is multimodal, accepting text, image, speech, and video input while producing text output, and it offers a large context window. It is available through a single API provider, and Google has not disclosed its parameter count. For readers tracking frontier model economics, this evaluation frames a capable but costly and talkative system.