llama.cpp adds local support for GLM 5 Flash MTP

A llama.cpp pull request adds support for GLM 5 Flash MTP, allowing users to run the model locally. The update was submitted by contributor pwilkin.

• Four open decision models benchmarked on an RTX 40902026-10-08
• User builds 256GB VRAM rig with 8 Radeon Pro V620s2026-10-08
• llama.cpp takes the stage2026-10-07
• llama.cpp Adds GPU Cache for MoE Experts in Host Memory2026-10-07
• llama.cpp v0.6.0 Released With MTP Decoding2026-10-05