MediaTek-Research/Breeze-Guard-26
TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:32kPublished:Mar 17, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

MediaTek-Research/Breeze-Guard-26 is an 8 billion parameter safety classifier developed by MediaTek Research, specifically designed for detecting harmful content in user prompts. Built on the Breeze 2 8B Instruct backbone, it is fine-tuned on 12,000 human-verified samples to identify Taiwan-specific safety risks. This model excels at prompt-level harmfulness detection in Taiwanese Mandarin, supporting six predefined risk categories including scam, financial malpractice, and political manipulation. It offers both thinking and non-thinking inference modes for explainability or low-latency applications.

Loading preview...