Home/Topics/prefix-cache Thread prefix-cache 1 run Signed runsThe measured evidence Model Tok/s Peak VRAM TTFT Verdict Qwen3.5-122B-A10B-MTP-GGUF · Q4_K_M NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition (96GB) · Single GPU 144.5 83 GiB 35 ms Verdict Great Mid-Size MoE