Is llama.cpp suitable and capable to run NVFP4 models
With Qwen3.8:27B arriving yesterday, including NVFP4 versions, can llama.cpp successfully run these models, or must I use a separate instance of vLLM for NVFP4 models specifically?
Searching on the internet initially made me believe that vLLM was the only way, but then some more recent posts hinted that llama.cpp is being developed for NVFP4 as well.
If llama.cpp is indeed capable of running NVFP4 models, I'd appreciate some pointers on how to get llama.cpp running NVFP4 models.