Google unveils Gemini 4 Carbon with enhanced coding, cybersecurity capabilities

ISTANBUL – Google has announced its latest artificial intelligence (AI) model, Gemini 4 Carbon, claiming significant improvements in coding, agentic tasks and cybersecurity defence, Anadolu Ajansi reported.

Unveiled just two weeks after the launch of its Gemini 4 Argon model, the company described Carbon in a blog post as its strongest frontier model, with advanced cybersecurity capabilities and the ability to handle complex, long-running tasks.

Google attributed the rapid release to its researchers and AI models "working hand in hand to advance capabilities faster".

The company said Gemini 4 Carbon ranked first on the updated version 5.0 of the Artificial Analysis Intelligence Index, scoring 59.

Based on user feedback from its previous release, Google said it had improved the model's performance in full-stack web development, scientific tasks and multimodal understanding. It also claimed the model had gained 89 points to rank fourth on the WebDev Arena leaderboard.

In cybersecurity benchmarks, Google said the model delivered leading results, scoring 88.2 per cent on SEC-bench Pro for identifying vulnerabilities, 54.8 per cent on ExploitGym for exploiting vulnerabilities and 71 per cent on CWE-bench for patching them.

"On internal benchmarks, Carbon drives a strong uplift in generating secure code, and in live coding environments with our own teams it shows reduced misalignment and fabrication rates," the company said.

On safety and alignment, Google said internal testing showed Carbon reduced unsupported claims by approximately half compared with Argon, while stating methodological limitations directly rather than fabricating missing evidence.

The company also said the model could retrieve relevant sources and highlight discrepancies when faced with ambiguity. It added that Carbon reduced task overreach by 26 per cent and under-service by 29 per cent during multi-step workflows.

Google said Gemini 4 Carbon would initially be made available to a limited group of cybersecurity defenders this week, before access was gradually expanded to developers, businesses and consumers.

The company said introductory pricing would be US$2 per million input tokens and US$10 per million output tokens, with cached input tokens discounted by 95 per cent. After the introductory period, prices would increase to US$4 per million input tokens and US$20 per million output tokens. –BERNAMA