BLAST RADIUS

Open-weight & owned stack · Quantization efficiency

llama.cpp Proposes Maple Ternary MoE Support

Sep 14, 2026

A llama.cpp pull request proposed CPU and low-VRAM support for DeepGrove’s preview Maple 20B-A1B ternary MoE architecture.

Read the original at reddit.comOpens the publisher's site in a new tab

More in Open-weight & owned stack