Breaking01 / AI
Google gives cyber defenders first access to Gemini 4 Carbon

Google DeepMind announced Gemini 4 Carbon on 28 Oct 2026 at $2 per million input tokens and $10 per million output tokens, with a subset of cyber defenders first in line for access.
The company said availability for developers, businesses and consumers comes after more work on guardrails, the safety controls built around an AI model.
Pricing. The announcement confirms that the rate is the same introductory price as Gemini 4 Argon, the company's earlier model. Cached input tokens are priced at 95% off the input token price.
Access. The company said access widens in phases to additional trusted partners through its Fairwind Program. It also said it is taking part in the US government's voluntary pre-release model testing process.
Carbon scored 88.2% on SEC-bench Pro, a vulnerability-finding benchmark, according to company-published figures. The same figures show 54.8% on ExploitGym, which tests exploitation, and 71.0% on CWE-bench, which tests patching. The company wrote that the three results were evaluated without guardrails.
Koray Kavukcuoglu, senior vice president at Google DeepMind, wrote in the Gemini 4 Carbon announcement that Carbon is "our strongest model". Carbon ranks first on Artificial Analysis Intelligence Index v5.0 with a score of 59, a company-published figure.
The company's comparison is with Argon. On internal benchmarks, Carbon's success rate in generating secure code was more than twice Argon's. The attack success rate on Gray Swan's indirect prompt injection benchmark, which measures attacks that hide instructions in content a model reads, fell (0.7% to 0.0%). Both results are company-published figures.
The company has not disclosed a date for availability to developers, businesses or consumers.