Google's latest frontier model, Gemini 4 Carbon, has demonstrated the ability to outperform leading models from OpenAI and Anthropic on specific performance metrics.
In a Wednesday announcement, Google revealed Gemini 4 Carbon, asserting that the model delivers "significant gains" in coding, agentic tasks, and cybersecurity. The company highlighted that Carbon currently holds the top position on the Artificial Analysis Intelligence Index (v5.0) benchmark, surpassing competitors like Claude Opus 5.5 and GPT-6 Astra.
This new iteration improves upon the previously released Argon model. While currently restricted to a select group of cybersecurity defense partners under the Fairwind Program, access will eventually expand to other collaborators.
Google describes Carbon as reflecting the rapid advancement of frontier AI, featuring a "significant leap" in cyberdefense capabilities. This upgrade enables the model to more effectively identify vulnerabilities and generate secure code. Currently, Carbon leads the rankings on both SEC-bench Pro, which evaluates a model's ability to discover real-world bugs, and ExploitGym, which tests the conversion of vulnerabilities into exploits.
Designed for complex, multi-step engineering tasks, Gemini 4 Carbon is capable of operating for extended periods to persist through obstacles. Google notes that it excels at avoiding shortcuts, resisting prompt injection attacks, and collaborating with users. Additionally, the model offers increased "transparency and honesty," reducing unsupported claims by half when compared to its predecessor, Argon.
The model will be introduced at a competitive introductory rate, priced at $2 per million input tokens and $10 per million output tokens.
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.