MBOX being a plain concatenation of messages with attachments MIME encoded inside, bulk extraction is genuinely scriptable, plus a couple of no code routes by taste:

The script route, most control: a short Python script using the standard library mailbox and email modules walks the MBOX, iterates each message's parts, and writes every part with a filename or a non text disposition to an output folder, deduplicating names as it goes. Twenty lines, no dependencies and it handles thousands of messages in seconds while letting you filter by extension, only PDFs and images per your need, the precision a GUI rarely offers.

The client route, no code: import the MBOX into Thunderbird with ImportExportTools NG, then the same add-on offers export attachments operations across a folder, writing them out in bulk. Slower to set up than a script if you lack Thunderbird, but entirely code free once it runs.

The converter route when the extraction is really a step toward something else: tools converting MBOX to PST or PDF often include an attachment extraction option, worthwhile when you also want the mail in another form, wasteful when attachments are genuinely all you want. Match the tool to whether attachments are the destination or a byproduct.

The detail worth handling whichever route: attachment name collisions, thousands of messages producing many image.png and scan.pdf names, resolved by prefixing each with its message date or a sequence number so nothing overwrites, the script doing this in a line and the GUI tools offering it as a naming option. Inline images embedded in message bodies versus genuine attachments is the other choice, most extraction wanting real attachments while filtering out the signature logos and tracking pixels that inline parts include.