Overview
Gemini 3.8 Flash Cyber is a post-trained version of Gemini 3.8 Flash designed for cybersecurity. Google positions it for trusted defenders, with a particular emphasis on autonomous vulnerability discovery and automated patching rather than exploitation. (docs.cloud.google.com)
Capabilities and uses
- Analyzes code for security vulnerabilities and assists with developing fixes. Google reports a 47.2% pass@1 result on the external CWE-Bench patching benchmark and over 70% success on an internal vulnerability-discovery benchmark spanning 20 programming languages; these are benchmark results, not guarantees of real-world performance. (blog.google)
- Accepts text, images, audio and video as input, and produces text output. The documented context window is 1,048,576 tokens, with a maximum of 65,536 output tokens. (docs.cloud.google.com)
- Supports thinking, system instructions, structured output, context caching, Google Search grounding, code execution and function calling. (docs.cloud.google.com)
Access and limitations
The model is generally available on Google Cloud behind an allowlist. Access can be requested through a Google team or the Fairwind Program, which prioritizes trusted government authorities, critical infrastructure operators and software maintainers. The model has more permissive cybersecurity mitigations than the general Flash variant. (docs.cloud.google.com)
Google Cloud lists batch inference, tuning, RAG Engine, URL context, computer use and the Gemini Live API as unsupported for this model. (docs.cloud.google.com)
Pricing
Google Cloud Gemini Enterprise Agent Platform prices, checked November 16, 2026, are in USD per 1 million tokens. Input covers text, image, video and audio; text output includes response and reasoning tokens. The same listed rates apply to the at-or-below 200,000 and above-200,000 input-token tiers. (cloud.google.com)
| Pay-as-you-go option | Region | Input | Cached input | Output |
|---|---|---|---|---|
| Standard | Global | $1.50 | $0.15 | $7.50 |
| Standard | Non-global | $1.65 | $0.165 | $8.25 |
| Priority | Global | $2.70 | $0.27 | $13.50 |
| Priority | Non-global | $2.97 | $0.297 | $14.85 |
| Flex | Global | $0.75 | $0.075 | $3.75 |
| Flex | Non-global | $0.825 | $0.0825 | $4.125 |
The pricing page groups Flex and batch rates together, but the model documentation says batch inference is unsupported; the Flex rates above should not be interpreted as confirmation of batch availability. Explicit cache storage is separately listed at $0.000001 per token per hour. Other platform, tool or service charges may apply. The announcement's introductory pricing for the non-Cyber Gemini 3.8 Flash is not assumed to apply to this model. (cloud.google.com)
Dates
- Announcement: September 2, 2026. (blog.google)
- Google Cloud model release: September 16, 2026, according to the model documentation. (docs.cloud.google.com)
Official sources
- Google announcement
- Google Cloud model documentation
- Google Cloud pricing — checked November 16, 2026.