Possible to load non shared experts to SSD in llama.cpp?

<!-- SC_OFF --><div class="md"><p>I have been using llama.cpp to run some models recently. For example, I've been running GLM-4.7-Flash with this command <code>.\llama-server.exe

<!-- SC_OFF --><div class="md"><p><a

<!-- SC_OFF --><div class="md"><p>If you are using a MoE model that does not fully fit in your GPU, some of the experts must stay on the CPU. Putting the experts that you will

<!-- SC_OFF --><div class="md"><p>So i recently updated LMstudio after a long pause and updated my llama.cpp runtimes too.. i was shocked.. i thought maybe something like

<!-- SC_OFF --><div class="md"><p>is it possible that a moe model can decide how many billion parameters to activate per token according to the task. eg if qwen 3.6 35b a3b - if a

<!-- SC_OFF --><div class="md"><p>I’ve been testing LM Studio on my RTX 5070 Ti (16GB) and Ryzen 9800X3D, running Unsloth Qwen3.5 35B (UD Q4_K_XL).</p> <p>Initially, I thought LM

<!-- SC_OFF --><div class="md"><p>currently in order to use MTP one needs to enable it in the starting argument of llama server.</p> <p>--spec-type draft-mtp<br />

<!-- SC_OFF --><div class="md"><p>Every output token the model generates requires reading the entire KV cache from High Bandwidth Memory.</p> <p>Here's what that actually looks