bug with built-in question answering model in mistral

## Describe the bug
Downloading the mistral 24B in the ai playground required downloading around 80 gigabytes of model files, which was not actually required at that time. Because mistralai/Mistral-Small-24B-Instruct-2501 has a consolidated model file on hugging face, and 10 separate model files, which add together to form the consolidated file, Therefore, it is actually only necessary to download 40G files, but the bug is that if the file is not enough for 80G, an error will occur.