Gemma4 doesn't work with default llama.cpp?

[cedarforestgiant]

Running Lemonade 10.2 on Mac and get the error:

Error preparing model: Failed to load model ‘Gemma-4-E2B-it-GGUF’: llama-server failed to start

Grabbed the latest llama.cpp b8779, and set in Terminal:

lemonade config set llamacpp.metal_bin="/Users/MyName/Downloads/llama-b8779/llama-server"

Still produces the same error. Lemonade backend still shows b8460.

How to update llama.cpp?