What is the difference between GENCODE and Ensembl annotation? That's actually the first question in the GENCODE FAQs:
The GENCODE annotation is made by merging the Havana manual gene annotation and the Ensembl automated gene annotation. ...In practical terms, the GENCODE annotation is identical to the Ensembl annotation.
I am looking at the mouse data for GENCODE M15 compared to Ensembl 90, which should be comparable according to both source. Total number of transcripts is 131,100 vs 131,195, so that difference is negligible. However, some subsets are very different. The number of protein-coding genes is 21,950 vs 22,598, which is a little more noticeable. Long non-coding RNA genes is 11,975 vs 8,980, so more than 30% drop. Thus, it seems like annotation is not really identical. Are those differences real or are they just counting the gene biotypes differently?