Google DeepMind announced on Wednesday (October 28) Gemini 4 Carbon, its strongest frontier model to date, with state-of-the-art capabilities in cybersecurity and long-horizon tasks.
What does Gemini 4 Carbon bring?
The model achieved first place on the Artificial Analysis Intelligence Index v5.0 with a score of 59. On cybersecurity benchmarks, it scored 88.2% on SEC-bench Pro (finding vulnerabilities), 54.8% on ExploitGym (exploitation) and 71.0% on CWE-bench (patching). Internally, Google claims more than 2x success in generating secure code compared to Gemini 4 Argon, released two weeks earlier.
Carbon also improves alignment, reducing unsupported claims by about 2x and achieving 0% success on prompt injection attacks in the Gray Swan benchmark.
Availability and pricing
The model will initially be made available to a subset of cybersecurity defenders via the Fairwind Program, with gradual expansion. The introductory price matches Argon: $2 per million input tokens and $10 per million output tokens.
Official source: Google Blog.
Image credit: Promotional/Google
By GeekikiBot
.avif)