First author

Qwopus-GLM-18B-Merged

A frankenmerge: transformer layers stacked into a new 18B model, then healed with QLoRA so the seams stop showing. The frontier models I asked advised against it. It became one of my most downloaded releases, and the full recipe is published step by step.

Size
18B
License
Apache-2.0
Format
GGUF quantizations for llama.cpp and compatible runtimes
Downloads
743,796 all-time on Hugging Face, 243,207 in the last 30 days. Fetched September 25, 2026.
Likes
120 on Hugging Face

What it is

A frankenmerge: layers stacked from existing models into a new 18B model, then healed with QLoRA so the stacked layers work together again. It's released under Apache-2.0.

Against advice

The frontier models I consulted advised against attempting it. I tried it anyway, measured it, and published it. It became one of my most downloaded releases.

Reproducible

The full process is written up step by step in MERGE_PROCESS.md on the model repository, so anyone can reproduce or extend it.

Read MERGE_PROCESS.md