WORLDTECHGoogle unveils Gemini 4 Carbon, first open to vetted cyber defenders
Google has introduced Gemini 4 Carbon as its next frontier model, only two weeks after releasing Gemini 4 Argon. The company says the model leads a recently updated benchmark and makes its largest advance in cybersecurity, which is why access is restricted at the outset.
Access limited to vetted defenders at first
Carbon will not reach the public at the start. According to Google, the model will first be made available to a subset of cyber defenders, with developers, businesses and consumers to follow later.
Pricing is set at $2 for every million input tokens and $10 for every million output tokens. Cached input tokens are charged at 95 percent below the standard input rate.
Gains in coding, agents and security
Google presents the model as designed for extended agent workflows and the entire cyber defense lifecycle. Besides agentic and scientific work, the company points to better full-stack web development and multimodal understanding.
Cybersecurity is where Google sees the biggest leap. Internal tests cited by the company indicate that Carbon generates more than twice as much secure code as its predecessor.
Benchmark placement and rivals
Google highlights Carbon's first place on the Artificial Analysis Intelligence Index (v5.0), where it scores 59. Claude Opus 5.5 and GPT-6 Astra are reported to sit behind it.
According to the announcement as covered by outlets reporting the launch, the model also reaches 88.2 percent on SEC-bench Pro.
Open questions
- When will Carbon reach developers, businesses and consumers beyond the initial group of defenders?
- How will Google's internal secure-code tests be verified independently?