Users are reporting significant reliability issues with the DeepSeek-V4-Flash-0731 model when performing non-coding tasks. Despite high intelligence benchmark scores, the model reportedly fails to grasp crucial subtleties and struggles with basic tasks. These inconsistencies make the model appear unreliable for general-purpose applications outside of specialized coding use cases.