FlashLabs releases uncensored Qwen3.8 model for security research

FlashLabs has announced the release of ‘Qwen3.8-Flash-Next-Uncensored’, an open-weight large language model designed specifically for security research. Developed by the US-based AI research firm Continuum AI and distributed exclusively in the Japanese market by FlashLabs, the model is based on Alibaba’s ‘Qwen3.8-Flash-Next’.

The uncensored version is intended to assist ‘red teams’ and ‘blue teams’ in identifying vulnerabilities by providing a model with relaxed content filtering, allowing researchers to observe behaviors that are often obscured by standard safety guardrails. The model supports multimodal inputs, including text, images, and video, and features a native context window of 262,144 tokens.

To prevent misuse, the model weights are available on Hugging Face via an application process and are subject to specific license terms restricting use to legitimate security research and verification. The release includes GGUF formats and native MLX formats optimized for Apple Silicon.