All Radar picks

X OPPORTUNITY RADAR / THE BRIEF

The Frontier Model Has a Guest List

Google’s strongest new AI model goes first to a vetted cyber-defender cohort. The capability claims are large; general access still has no firm date.

Information Overload
1 min readSources
Conceptual editorial illustration of a black server core behind a glass security checkpoint, with a small group of cyber defenders admitted while a larger developer crowd waits outside.
AI-generated conceptual editorial illustration; not documentary evidence.

Google’s strongest new model is arriving first through a cybersecurity checkpoint. Gemini 4 Carbon will initially go to a subset of defenders in the Fairwind Program and then to trusted partners. Broader access is promised, but Google gave no firm public date.[1]

Google reports strong cyber results: 88.2% on SEC-bench Pro vulnerability finding, 54.8% on ExploitGym exploitation and 71.0% on CWE-bench patching. Several numbers are internal, and the company says the cyber evaluations were run without guardrails. Treat them as company claims awaiting independent reproduction.[1]

The capability has a guest list.

The rollout matters because access is becoming part of AI governance. The same company building and pricing the capability is deciding who gets the first look. Safety may justify a controlled release; it does not make the gate apolitical.

Introductory API pricing is $2 per million input tokens and $10 per million output tokens, later scheduled to double. The public can inspect the tariff before most of it can inspect the model. A very modern form of transparency.[1]

FOLLOW THE SOURCES.

Checked Oct 28, 2026. Numbered references point to the evidence behind factual passages. Commentary is our own.

  1. Introducing Gemini 4 Carbon, our most capable and intelligent model yet — October 28, 2026 (opens in a new tab)GoogleBack to cited passage
AICybersecurityTechnology marketsAI governanceGoogle
Back to the latest picks