Mountain View:Google has announced the launch of its latest artificial intelligence model, Gemini 4 Carbon, claiming significant advancements in coding, autonomous agent tasks, and cybersecurity defense. This new model follows the recent release of Gemini 4 Argon and is described as the company's strongest frontier model, featuring state-of-the-art cybersecurity and long-horizon capabilities.
According to Anadolu Agency – English, Google attributed the rapid timeline for this release to the collaboration between its researchers and AI models, which has accelerated the development of new capabilities. The Gemini 4 Carbon has achieved the top spot on the updated version 5.0 of the Artificial Analysis Intelligence Index, with a score of 59.
The model incorporates improvements based on user feedback from previous releases, enhancing full-stack web development, scientific tasks, and multimodal understanding. This has resulted in an 89-point leap to fourth place on the WebDev Arena leaderboard. In cybersecurity benchmarks, Gemini 4 Carbon leads with an 88.2% score on SEC-bench Pro for identifying vulnerabilities, 54.8% on ExploitGym for exploitation, and 71% on CWE-bench for patching.
Google has noted that internal testing shows the model significantly improves the generation of secure code and reduces misalignment and fabrication rates in live coding environments. The model also reportedly halves unsupported claims compared to its predecessor, Argon, and retrieves relevant sources while highlighting discrepancies to the users. It reduces task overreach by 26% and under-service by 29% during multi-step workflows.
Initially, Gemini 4 Carbon will be available to cybersecurity defenders before becoming accessible to developers, businesses, and consumers. Google plans to offer introductory pricing for the model at $2 per 1 million input tokens and $10 per 1 million output tokens, with a 95% discount on cached input tokens. Standard pricing will later increase to $4 per 1 million input tokens and $20 per 1 million output tokens.