alt Hacker Newssome support already merged, and I verified in a local build that it runs (cannot get MTP params working tho, about ~40 tok/s on my beefy 800GB/s 7900XT w/ 20GB VRAM). https://github.com/ggml-org/llama.cpp/pull/26841
some support already merged, and I verified in a local build that it runs (cannot get MTP params working tho, about ~40 tok/s on my beefy 800GB/s 7900XT w/ 20GB VRAM). https://github.com/ggml-org/llama.cpp/pull/26841
Cant wait to kick the tires. I am running qwen3.5-coder:35b on my laptop, and while thats nice, such alternatives are welcome.