Yes, and they specifically mention "Up to 10.7x faster LLM prompt processing in LM Studio" which is probably using the neural accelerator for prefill.