GLM-5.3-Flash graftB Q6_K · MTP · ik_llama.cpp 24.2 tok/s GLM-5.3-Flash-graftB.gguf ik_llama.cpp425a2c1dQ6_Krtx-a6000rtx-3090linux1 stream! 2 caveats midagedev 2026-09-18
Offload ladder · 10 layers on CPU 102 tok/s Qwen3.6-35B-A3B-UD-Q6_K.gguf ik_llama.cppc10fbbccUD-Q6_Krtx-a6000linux1 stream! 2 caveats midagedev 2026-09-18
Offload ladder · 0 layers on CPU 129 tok/s Qwen3.6-35B-A3B-UD-Q6_K.gguf ik_llama.cppc10fbbccUD-Q6_Krtx-a6000linux1 stream! 3 caveats midagedev 2026-09-18
Qwen3.6-35B-A3B Q6_K · 1 stream · ik_llama.cpp 129 tok/s Qwen3.6-35B-A3B-UD-Q6_K.gguf ik_llama.cppc10fbbccUD-Q6_Krtx-a6000linux1 stream! 3 caveats midagedev 2026-09-18