---
license: apache-2.0
language:
  - en
  - zh
library_name: gguf
pipeline_tag: text-generation
tags:
  - minicpm
  - minicpm5
  - llama
  - text-generation
  - on-device
  - edge-ai
base_model: openbmb/MiniCPM5-1B
---

# MiniCPM5-1B-GGUF (Q4_K_M)

Mirror of [openbmb/MiniCPM5-1B-GGUF](https://huggingface.co/openbmb/MiniCPM5-1B-GGUF)'s
`MiniCPM5-1B-Q4_K_M.gguf`. Used by the IceSpiritAI_Chat Android app (MiniCPM5-1B
GGUF backend via llama.cpp; alternative to the default Qwen3.5-2B-MNN LLM).

## Identity

| Field | Value |
| --- | --- |
| Source | `huggingface.co/openbmb/MiniCPM5-1B-GGUF` (official) |
| File | `MiniCPM5-1B-Q4_K_M.gguf` |
| Size | 688,065,920 bytes (656.30 MiB) |
| SHA-256 | `81b64d05a23b17b34c475f42b3e72fbde62d4b92cc34541f7a8031d0752deafa` |
| Architecture | Standard `LlamaForCausalLM` (per OpenBMB model card) |
| Params | 1.08B (24 layers, GQA 16+2, ctx 131072) |
| Tokenizer | `gpt2` (llama-bpe pre-tokenizer) |
| Uploaded | 2026-06-25 |

## Why this mirror exists

IceSpiritAI_Chat is a dual-LLM Android app. The default LLM is
`Qwen3.5-2B-MNN` (small, fast, on-device MNN); the alternative is
`MiniCPM5-1B-GGUF` (slightly larger, higher-quality generations, served by a
llama.cpp native pipeline). Users in mainland China without reliable access to
`huggingface.co` can use this mirror or the ModelScope mirror
[`AlexZh/MiniCPM5-1B-GGUF`](https://modelscope.cn/models/AlexZh/MiniCPM5-1B-GGUF).