Abstract-The fast advancement in 6G communication technology increases the scalability, intelligence, and connectivity of the IoT ecosystem. Nevertheless, the emergence of heterogeneous IoT devices, ultra-low latency communication, and decentralized network infrastructure makes 6G infrastructure vulnerable to advanced cyber threats, which the existing intrusion detection techniques fail to detect efficiently. Machine learning and deep learning techniques are often constrained by their limited understanding of the context, lack of proper modeling of complex communication relationships, and inability to learn new patterns of attacks leading to poor threat detection accuracy and delayed responses to cyber threats. To overcome these challenges, this paper presents a Graph Attention based Cyber Threat Intelligence Framework (GACTI-6G), which provides an intelligent solution to intrusion detection and adaptive response to threats in 6G IoT ecosystems. This paper uses network entities as nodes in the graph and communication behavior between them as edges for capturing relational dependencies using multi-head Graph Attention Network (GAT). The Cyber Threat Intelligence Fusion module enriches contextual information using network flow characteristics, threat signatures, and behavioral characteristics. An adaptive response engine prioritizes the detected threats based on severity scores and provides recommendations for mitigating threats. Experiments have been conducted, and the results indicate that the accuracy, precision, recall, and F1-score obtained from the use of the proposed GACTI-6G framework are 98.31%, 97.82%, 97.25%, and 97.54% respectively, where the average detection latency is reduced to 12.9 milliseconds. The comparison with other intrusion detection techniques, including those based on CNN, LSTM, and Graph Attention Networks, proves the superiority of the proposed method in offering highly accurate, scalable, and real-time cyber threat intelligence for 6G IoT networks.