Did LLMs finally subsume embedding models? We compared 10 LLMs vs. 26 embedding models across 37 MTEB tasks. LLMs scored 77.6 while embedding models scored 77.2 But the costs are very different. 🧵