Mbili is two. I learn phonetic by Wikipedia, so I could see how that b-v happens. But, I couldn't see why the b would grammatically do so, or is it just some normal phenomenon?
"Mbili" is only used when counting abstractly or when agreeing with nouns in the N class, such as chupa mbili (two bottles), dakika mbili (two minutes), etc.
The root word for the numeral two is actually -wili, and it takes adjective agreements based on the noun it modifies. For the Ki/Vi class, the plural agreement is vi- so you get viwili, as in viatu viwili (two shoes), visu viwili (two knives), etc. Another example is the Ma class, in which the plural agreement is ma- so you get mawili, as in makanisa mawili (two churches), majibu mawili (two answers), etc.
Note that -wili only takes plural agreements since of course we are talking about two of something!
is this true for all numbers? or do some numbers stay the same everytime (ex: saba, moja) - and if do, then how do we know which ones?
Good question. The numbers 1, 2, 3, 4, 5 and 8 are all of Bantu origin and so take agreements (-moja, -wili, -tatu, -nne, -tano and -nane). The numbers 6, 7 and 9 are all of Arabic origin and so do not take agreements (sita, saba and tisa).
The root form is -wili. In the N class (9/10), the nasal consonant added to the beginning depends on and affects the first consonant. W changes to mb- in this situation.