An unexpected behavior has been observed in the Claude Sonnet-4.6 language model: when prompted in Chinese, the model responds by self-identifying as DeepSeek-V3.

Details of the Anomaly

The discovery was initially shared on X (formerly Twitter) by user TeortaxesTex and subsequently discussed on the Reddit platform. The attached image shows the model's response stating that it is DeepSeek-V3.

Possible Implications

This behavior raises questions about the model's internal identification mechanisms and its architecture. It may indicate an internal confusion between models or an error in handling requests in different languages. Further analysis is needed to fully understand the cause of this phenomenon.

For those evaluating on-premise deployments, there are trade-offs to consider. AI-RADAR offers analytical frameworks on /llm-onpremise to evaluate these aspects.