Traffic analysis and classification in O-RAN support control decisions such as scheduling and QoS enforcement. Recent work applies AI models to improve classification accuracy. However, different models introduce different inference latencies, which affect how quickly the network can react. We present a demo that enables direct observation of how AI models impact the performance of traffic analysis and classification in O-RAN. We deploy multiple models as xApps and evaluate them under the same traffic conditions. Users can select a model and observe classification accuracy and inference latency in real time, revealing the trade-off between accuracy and response time. The demo is implemented using OpenAirInterface (OAI) with a Near-RT RIC and can be extended to other O-RAN deployments.