MOOCs (massive open online courses) are an online teaching proposal that, in their short lives, have already thrown up two very different possibilities: cMOOCs and xMOOCs. Both are analysed in this paper from the perspective of assessing students’ learning. While assessment in xMOOCs is usually limited to multiple choice tests and sometimes delivering tasks, in cMOOCs the aim is to foster interaction from an educational standpoint, usually on the basis of peer assessment. Pedagogically, both models have their limitations. Multiple choice tests are mainly content bound while peer assessment has its own difficulties and drawbacks, which we explain here. We will put forward some useful ideas to give more flexibility to assessment in MOOCs (groups of experts, semantic web, portfolio, learning analytics), in an attempt to address educational assessment not as an end in itself but as another part of the educational process.