Google DeepMind Announces Gemini 4 Carbon
Google DeepMind has unveiled its latest frontier model, Gemini 4 Carbon, highlighting state-of-the-art cybersecurity and long-horizon reasoning capabilities.
Radwan AbdulhadiEditor and technology writer
Introduction to Gemini 4 Carbon
Google DeepMind has officially announced its next frontier model, Gemini 4 Carbon, positioning it as the company's strongest model to date with advanced cybersecurity and long-horizon capabilities. Building directly upon the release of Gemini 4 Argon just two weeks prior, the new model demonstrates the rapid pace at which researchers are pushing technological boundaries while simultaneously enhancing safeguards.
According to the company, Carbon ranks number one on the updated version 5.0 of the Artificial Analysis Intelligence Index with a score of 59. The system exhibits substantial improvements across coding, agentic operations, and critical cybersecurity defense tasks.
Long-Horizon Engineering and Reasoning
Engineered to tackle complex and sustained engineering tasks, Gemini 4 Carbon is built for multi-step agentic workflows. Internal team evaluations have demonstrated that the model can run for extended periods to resolve intricate problems, root-cause issues, and patch subtle bugs across massive codebases and low-level software systems.
Responding directly to user feedback, Google DeepMind concentrated on improving performance across scientific tasks, full-stack web development, and multimodal understanding. Improvements in output quality and artifacts have propelled Carbon upward on the WebDev Arena leaderboard.
Cybersecurity and Vulnerability Management
Gemini 4 Carbon achieves significant performance leaps across the entire cybersecurity defense lifecycle, covering vulnerability detection, validation, and remediation. The model secured leading positions on multiple benchmarks, including SEC-bench Pro for identifying software flaws, ExploitGym for exploitation testing, and CWE-bench for root-cause patching.
Furthermore, cloud security platform Wiz observed that Carbon excels in benchmarks evaluating web penetration testing, which serves as a core capability for its Scan for Good initiative.
Safety, Alignment, and Rollout Strategy
The new frontier model incorporates a multi-layered safety stack combining model training, prompt-injection resilience, real-time misuse defenses, and chain-of-thought monitoring. In evaluations, Carbon achieved a zero percent success rate on indirect prompt injection resistance benchmarks and substantially reduced shortcut attempts and unsupported claims during internal testing.
Google DeepMind is pursuing a deliberate rollout strategy by participating in the United States government's voluntary pre-release testing process. Initial access is being extended to select cyber defenders and trusted partners through the Fairwind Program before expanding availability to developers, businesses, and consumers.
Pricing and Availability
Gemini 4 Carbon launches at the same introductory price as Argon, charging $2 per million input tokens and $10 per million output tokens, with cached input tokens discounted at 95% off the standard input rate. Following the introductory period, standard pricing will adjust to $4 per million input tokens and $20 per million output tokens.
Sources
- Google DeepMindGemini 4 Carbon: Our next frontier model
Continue chronologically
Related entity coverage
Related stories
Google Announces Gemini 4 Carbon Frontier AI Model
Google has announced Gemini 4 Carbon, its latest frontier AI model featuring performance improvements over Argon, though it is not yet widely available.
Google Introduces Gemini 4 Carbon for Cyber Defense
Google has announced Gemini 4 Carbon, a new frontier AI model designed to find, exploit, and patch software vulnerabilities with advanced capabilities.
Google DeepMind Unveils Gemini 4 Argon Frontier AI Model
Google DeepMind announces Gemini 4 Argon, a new frontier AI model designed for real-world coding, enterprise workflows, and cyber defense.
Google DeepMind Details Gemini 4 Argon Safety and Security
Google DeepMind outlines safety, security, and mitigation measures for Gemini 4 Argon, addressing misuse, alignment, and real-time safeguards.
