I’ve observed some inconsistencies between the model used by the Assistance API and the GPT Plus model. Occasionally, it generates information that isn’t included in the provided data (files). Both my custom GPT model and the Assistant API are configured with identical instructions and files. However, while the custom GPT yields accurate responses in line with my data, the Assistant API does not. This discrepancy persists across both the gpt-4-1106-preview with the retrieval tool and the GPT-4 model with the code interpreter tool.