# llama: add Maple 20B-A1B ternary MoE architecture (CPU) by AlexGabbia · Pull Request #27000 · ggml-org/llama.cpp — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/904eaa0c50e92a0f

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-09-14 12:11 ingested 2026-09-15 02:40

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1wg1o5b/llama_add_maple_20ba1b_ternary_moe_architecture/)

## retained text



```
20B-A1B model is coming, good for low VRAM people?

[https://huggingface.co/deepgrove/maple-preview](https://huggingface.co/deepgrove/maple-preview)
```



document fetched 2026-09-15 00:43 · canonical https://reddit.com/r/LocalLLaMA/comments/1wg1o5b/llama_add_maple_20ba1b_ternary_moe_architecture
