Google Announces Gemini 4 Carbon

Google has announced Gemini 4 Carbon, a frontier AI model featuring cybersecurity capabilities and designed for long-horizon engineering tasks. Coming two weeks after the launch of Gemini 4 Argon, the model ranks #1 on v5.0 of the Artificial Analysis Intelligence Index with a score of 59.

The model is built to sustain agentic workflows across multi-step engineering problems, with internal testing showing it running for days to patch software bugs across codebases. On the WebDev Arena leaderboard, Carbon moved up +89 elo points to reach #4 with a score of 1768.

In cybersecurity evaluations, Carbon scored 88.2% on SEC-bench Pro for identifying vulnerabilities, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for root-cause patching. Internally, Google reported a more than 2x uplift in generating secure code compared to Argon, while its Chrome team observed over 10X better performance than another lab's large frontier model when generating patches.

Google also detailed safety and alignment updates in Carbon:

• Successful cheating on Terminal-Bench 4.0 fell to 0.0%, with cheat attempts dropping to 0.4%.
• High-severity misalignment alerts in internal live coding environments dropped by almost 80%.
• Unsupported claims decreased by approximately 2x compared to Argon.
• Overreach dropped by 26% and under-service declined by 29%.
• The model achieved a 0.0% attack success rate on Gray Swan's Indirect Prompt Injection Benchmark, down from 0.7% with Argon.

Google is rolling out access starting this week to select cyber defenders and trusted partners through its Fairwind Program while participating in the U.S. government's voluntary process for pre-release testing. Wider availability for developers, businesses, and consumers will follow. Carbon launches at an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens discounted by 95%. After the introductory period, pricing shifts to $4 per 1M input tokens and $20 per 1M output tokens.