Google has launched Gemini 4 Carbon, a frontier artificial intelligence model it describes as substantially more capable than the one it shipped less than a month earlier, according to The Information.
The launch caps nearly a year in which Google had not released a new flagship model, making the pace of the past month a marked departure from its prior cadence.
Carbon arrives roughly two weeks after Gemini 4 Argon, which Google began rolling out globally to Gemini app subscribers on 13 October as a model for completing complex, multi-step tasks such as financial analysis and presentations.
Carbon is pitched differently: Google calls it its strongest frontier model yet, built for cybersecurity, vulnerability discovery, exploitation, secure patching and extended software-engineering work.
Access is initially limited to a subset of cyber defenders, with trusted partners to follow through Google's Fairwind Program; no date has been set for wider developer, business or consumer availability.
Introductory pricing is set at $2 per million input tokens and $10 per million output tokens, with cached input tokens discounted by 95%.
The staged rollout reflects Google's stated approach of testing safeguards and gathering feedback before opening access to more autonomous systems, as it works to strengthen its standing against OpenAI and Anthropic in coding, agentic workflows and cybersecurity.
Read more about Google
