Google's Gemini 4 Argon is now generally available and Gemini 4 Carbon was announced

Google released the Gemini 4 family in October 2026, closing the month with the general availability of Gemini 4 Argon and then announcing Gemini 4 Carbon two weeks later. Gemini 4 Argon is the newest frontier reasoning model from Google, which said it excels at deep work on complex, real-world problems in areas such as coding, finance, legal analysis, cybersecurity and scientific research. Google reported that Argon ranked first on the Vals Index across critical economic sectors, including finance, coding, legal work and tax, and priced the model at 2 dollars per million input tokens and 10 dollars per million output tokens, with cached input tokens at 95 percent off. Google announced Gemini 4 Carbon on October 28 as its strongest model yet, focused on long-horizon engineering and cybersecurity defense. Google reported that Carbon reached the top position on the Artificial Analysis Intelligence Index version 5.0 with a score of 59, and cited results of 88.2 percent on SEC-bench Pro, 54.8 percent on ExploitGym and 71.0 percent on CWE-bench, describing improvements in generating secure code and patching. Google describes both releases as backed by a multi-layer safety stack that includes jailbreak resistance, agent alignment monitoring and prompt-injection defenses. Carbon is designed for sustained agentic workflows lasting several days, Google said, and will start with pre-release testing through the Fairwind Program before wider access. The fast move from Argon to Carbon shows how quickly Google is advancing frontier AI capabilities. The general availability of Argon to broader developer and enterprise audiences, and the early reported results for Carbon, point to longer agentic coding and security workflows in the near term. Source: [Google](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/) ~AI