IB papers are given grades from 7 (highest grade) to a 1 (the lowest grade). The mark boundaries for each grade is different for each subject and for each level in a subject. The construction of the grade boundaries is quite a complex and detailed  procedure. 

The grade boundaries for each subject for the last number of years can be found here. 

Once the papers are all submitted to the IB they are scanned and marked. There is a grade awarding conference for each subject where the grade descriptors for each subject are compared to the answers of the students and viewed alongside the marks of the students. In some exam system a fixed percentage of students will achieve the same grade every year - this is called norm-referencing. This is not the main consideration for the awarding of grades in the IBDP.  The grade descriptors are more important than the percentage of the cohort achieving a certain mark. 

“The IB uses an approach known as weak criterion referencing, which is based upon criterion but recognises the evidence of the Good and Cresswell effect (where expert judgment does not accurately take into account the demand of the questions.). In this approach, expert examiners are asked to establish a narrow range over which the grade boundaries could lie based on the criterion (grade descriptors) and this is then compared with boundaries calculated to match performance from previous years. Where these two boundaries align the grades are set, but if they disagree there is further discussion to establish how this contradictory evidence can be aligned.” -Assessment principles and practices—Quality assessments in a digital age - IBO 2018