Two weeks after Gemini 4 Argon, Google introduces Gemini 4 Carbon as its most powerful model. According to the company, it ranks first in version 5.0 of the Artificial Analysis index, with 59 points.

The focus is on cybersecurity. Google’s figures:

• SEC-bench Pro (vulnerability detection): 88.2%

• ExploitGym (exploitation): 54.8%

• CWE-bench (remediation): 71.0%

• Indirect prompt injection (Gray Swan): 0.0% of successful attacks, compared with 0.7% for Argon

Google also says its internal teams have seen Carbon work for days on multi-stage engineering tasks, such as identifying the cause of subtle bugs in large codebases.

The rollout is phased. First, it reaches a group of cybersecurity experts and then, gradually, trusted partners through the Fairwind program. Google is also participating in the U.S. government’s voluntary pre-launch testing. The starting price is $2 per million input tokens and $10 per million output tokens, the same as Argon.

More information at territorioblockchain.com

#IA #Google #Cybersecurity