GLM-5.3-Flash-Uncensored-FP8: FP8 Quantized and Ablated Model Based on GLM-5.3 Flash

This is a modified model of Zhipu AI's GLM-5.3 Flash, quantized to FP8 precision and applying ablation techniques to remove safety filters. While retaining the original Mixture of Experts architecture and Vision-Language capabilities, it provides an unrestricted response environment required for AI Red Teaming and security testing.

It supports multimodal inference that processes image inputs in addition to text generation. By supporting Function Calling and Reasoning Capabilities, it maintains the same interaction methods as the original model in scenarios requiring complex agent workflows or logical reasoning.

It supports both English and Chinese and is distributed under the MIT License. It is designed to allow security researchers and red teaming experts, who need to bypass content restrictions while maintaining the performance of the existing GLM series, to experiment on a verified open-source foundation.