Google designed the new system to tackle complex coding and defensive operations, including finding vulnerabilities and generating software patches. Internal testing demonstrates that the model produces safe code at more than twice the success rate of the preceding Argon release. Access remains restricted to a select group of defenders under the Fairwind Program to evaluate safeguards before a broader launch. When standard availability opens, the company plans to charge $2 per million input tokens and $10 per million output tokens.
Key points
Google is rapidly expanding its Gemini model family to capture developer and enterprise m…: Google is rapidly expanding its Gemini model family to capture developer and enterprise market share from key industry rivals.
The company is prioritizing specialized security functions such as vulnerability discover…: The company is prioritizing specialized security functions such as vulnerability discovery, software exploitation, and automated patching.
Enterprise customers will eventually pay the same introductory pricing structure establis…: Enterprise customers will eventually pay the same introductory pricing structure established for the previous Argon release.
Sources checked by our analysis pipeline: 1
Oct 28 : Alphabet's Google on Wednesday unveiled Gemini 4 Carbon, its latest frontier AI model, stepping up competition with OpenAI and Anthropic as the race to build more capable coding and cybersecurity systems intensifies.
The launch comes two weeks after Google introduced Gemini 4 Argon, as the company expands its Gemini family of AI models for developers and enterprise customers.
Here are some details:
• The company said the Carbon model delivers state-of-the-art performance across cybersecurity tasks including vulnerability discovery, exploitation, and software patching
• Google said Carbon generates secure code at more than twice the success rate of Gemini 4 Argon on internal benchmarks
• The model is designed to handle complex, long-running engineering and coding tasks
• Carbon is being made available first to a subset of cybersecurity defenders through Google's Fairwind Program as the company continues testing safeguards
• Gemini 4 Carbon will be priced at $2 per million input tokens and $10 per million output tokens, matching Gemini 4 Argon's introductory pricing
Google has unveiled its latest frontier AI model designed for sustained deep reasoning across demanding long-horizon workflows. Initial access is restricted to vetted cybersecurity defenders and U.S. government evaluation programs prior to a broader public and enterprise launch. Internally, the architecture is already delivering substantial infrastructure optimizations, including migrating legacy systems code and reducing data center memory footprints.
Federal antitrust regulators are examining whether the massive transaction will unfairly squeeze competing developers across the broader artificial intelligence sector. Hugging Face serves as critical open-source infrastructure for millions of researchers, raising concerns about market consolidation as the chip giant expands its footprint. This regulatory check marks another hurdle for the hardware maker's aggressive software and developer ecosystem acquisition strategy.
The tech giant's latest frontier release aims to outpace top industry competitors while reinforcing critical software defenses against prompt injection and complex vulnerabilities. By targeting multi-day engineering workflows and reducing unverified claims, the platform provides specialized automated reasoning for enterprise development teams. Its deployment marks an aggressive push to lead high-stakes developer benchmarks without increasing token usage costs.