It should be bytes.
I linked the org-wide call as it's simpler, but have not used it for ages, TBH I also use...
https://developer.cisco.com/meraki/api-v1/get-network-appliance-uplinks-usage-history/
...as the time series data is more useful.
When setting up, I tested API figures with GB amounts of data going through lab devices, it matched reasonably well. I never bother looking at the UI data as it's not how we monitor/analyse, also see comment below. I may do another controlled test sometime as a check to see if anything odd is going on with API numbers.
I'm always wary of comparing Dashboard UI numbers with API numbers. Unless you know the timing and rules used by Dashboard exactly match what you do with the API, there's no reliable comparison. Also, I suspect some Dashboard values may be cached results from a periodic calculation, not gathered and calculated on the fly when the page is rendered.