Model
Model explorer

Gemini 3.8 Flash Cyber

CLOSED
Google · Gemini 3.8 family · released Sep 2, 2026

Released alongside Gemini 3.8 Flash on September 2, 2026 for trusted defenders through Google's new Fairwind Program — one core model, two access envelopes. Google reports frontier-level autonomous vulnerability discovery on CyberGym, surpassing both 3.5 Flash Cyber and significantly larger frontier models, plus an internal benchmark spanning 20+ languages built because CyberGym is C/C++-only; neither carries a published numeral, so neither is recorded. The one citable number is CWE-Bench patching. Wiz separately reports +7.5-9.7% higher recall on its internal penetration-testing benchmark at 2.3-5.2x lower cost than other leading frontier models. No public per-token rate — access is program-gated, not self-serve — so the entry ships with a single result row and stays below the ranking-coverage floor.

ReasoningCodingVisionFunction callingTool useAgentic
1000.0
Elo · unrated
Parameters
Undisclosed
Active params
Undisclosed
Context
1M tokens
Architecture
The Gemini 3.8 Flash core specialised for defensive cybersecurity, served under a separate access envelope; same Flash-tier speed and cost profile
License
Proprietary (Fairwind Program trusted-defender access)
Languages
API price (in/out)
No hosted API
Modalities
text · vision · audio · video
Benchmark results
Bar shows position within the tracked field; marker = field best
CWE-BenchCoding47.2%#1
best: this model · 47.2%
Run it locally
Closed weights — available via API only. No local deployment.
Input / M tok
Output / M tok
Gemini 3.8 family
Elo progression across releases
API price · each benchmark row carries its own source badge (see methodology)