Google launches Gemini 3.8 Flash with a cybersecurity-focused variant

Google has launched Gemini 3.8 Flash alongside Gemini 3.8 Flash Cyber, adding a security-focused option to its rapidly expanding AI model family.

The company positions Gemini 3.8 Flash for software engineering, agentic workflows and multi-step reasoning. Google lists introductory API pricing of $0.75 per million input tokens and $3.75 per million output tokens.

A model with a security job

Gemini 3.8 Flash Cyber is being offered to trusted defenders through Google’s Fairwind Program rather than as a general public release. Google says the model is aimed at finding vulnerabilities and helping with defensive security work.

Google reports scores of 73.7% on DeepSWE v1.1 and 54.9% on HLE-Verified for Gemini 3.8 Flash. It also says the Cyber model exceeded 70% on an internal vulnerability-discovery benchmark and achieved 47.2% pass@1 on CWE-Bench patching.

Those figures are company-provided results, not independent testing. Google has not supplied a complete comparison methodology for every benchmark, and Fairwind eligibility and regional availability remain limited or unclear.

Our opinion

Gemini 3.8 Flash Cyber is an interesting sign that AI vendors are carving out specialist security models rather than treating every task as a generic chatbot problem. The pricing looks competitive on paper, but the benchmark claims need outside scrutiny and the Cyber variant is not a normal public launch.