Obviously. Because they didn't train the model on proprietary MS code. Which is bad but also good in some way, as it might force MS to support better their formats in the open source world.