Arabic has had two systems of numeral notation in use at once for most of its history, and they came from different places. The older is the abjad reckoning, in which the letters of the alphabet carry values fixed by their position in the inherited Semitic sequence: kaf is twenty because it is the eleventh letter of that sequence. A number is written by setting down the letters whose values sum to it, largest first, so the notation is additive and has no zero and no place value.
The other is the set of figures now used everywhere and called, in English, the Arabic numerals. They are not Arabic in origin. They came into Arabic from India, and Arabic writers were clear about it: the technique is called Indian reckoning in the mathematical literature, and the treatise of al-Khwarizmi that carried it into general use in the Islamic world and, in Latin translation, into Europe, is named for that origin. The system has place value and a zero, which is what makes written arithmetic possible, and it displaced the abjad for calculation.
The abjad did not disappear. It survived for exactly the things place-value notation is bad at: numbering the sections of a book, marking verse groups in a manuscript, and above all composing chronograms, in which a phrase is written so that the values of its letters sum to the year of an event, so that a building or a book carries its own date in words. Two shapes of the modern figures are in use as well, the eastern forms used from Egypt eastward and the western forms that Europe adopted, which is a further reason a number on an Arabic page has to be read against the tradition it was written in.