Instructions to use openbmb/MiniCPM-o-4_5-gguf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use openbmb/MiniCPM-o-4_5-gguf with Transformers:

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("openbmb/MiniCPM-o-4_5-gguf", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use openbmb/MiniCPM-o-4_5-gguf with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf openbmb/MiniCPM-o-4_5-gguf:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf openbmb/MiniCPM-o-4_5-gguf:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf openbmb/MiniCPM-o-4_5-gguf:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf openbmb/MiniCPM-o-4_5-gguf:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf openbmb/MiniCPM-o-4_5-gguf:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf openbmb/MiniCPM-o-4_5-gguf:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf openbmb/MiniCPM-o-4_5-gguf:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf openbmb/MiniCPM-o-4_5-gguf:Q4_K_M

Use Docker

docker model run hf.co/openbmb/MiniCPM-o-4_5-gguf:Q4_K_M

LM Studio
Jan
Ollama
How to use openbmb/MiniCPM-o-4_5-gguf with Ollama:
```
ollama run hf.co/openbmb/MiniCPM-o-4_5-gguf:Q4_K_M
```
Unsloth Desktop

How to use openbmb/MiniCPM-o-4_5-gguf with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf openbmb/MiniCPM-o-4_5-gguf:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "openbmb/MiniCPM-o-4_5-gguf:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use openbmb/MiniCPM-o-4_5-gguf with Docker Model Runner:
```
docker model run hf.co/openbmb/MiniCPM-o-4_5-gguf:Q4_K_M
```

Lemonade

How to use openbmb/MiniCPM-o-4_5-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull openbmb/MiniCPM-o-4_5-gguf:Q4_K_M

Run and chat with the model

lemonade run user.MiniCPM-o-4_5-gguf-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use openbmb/MiniCPM-o-4_5-gguf with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf openbmb/MiniCPM-o-4_5-gguf:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default openbmb/MiniCPM-o-4_5-gguf:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use openbmb/MiniCPM-o-4_5-gguf with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf openbmb/MiniCPM-o-4_5-gguf:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "openbmb/MiniCPM-o-4_5-gguf:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

why it's painfully slow on RTX with 12gb Vram?

by WhatABigWorld - opened 7 days ago

Discussion

WhatABigWorld

7 days ago

I found the windows installer which saved my day "Comni-v1.0.22-win64.exe"
I downloaded
the MiniCPM-o-4_5-Q4_K_M.gguf and the other folder except for the weights folder which I think it's apple related
the Comni reads all tts llm vision and detcted my gpu rtx 12gb fine but when I start the server button it takes 4 to 5 minutes to load
if it's only for the first time that's gonna be fine for me but after I use the user interface and I use any of the 4 categories it's paniful it take ages to respond or to move from state to another like from start to live.
and if it respond it might do it once then it give silence for ever
I also noticed that this whole setup utlize only 13% of my gpu usage or none , same on cpu
in contrary I run llms on the same gpu much much faster so what's going ?

WhatABigWorld changed discussion title from why it's paninfully slow on 12g Vram? to why it's painfully slow on 12gb Vram? 7 days ago

WhatABigWorld changed discussion title from why it's painfully slow on 12gb Vram? to why it's painfully slow on RTX with 12gb Vram? 7 days ago

tc-mb

OpenBMB org 6 days ago

Thanks for the report. Skipping weights/ is fine — that path is Apple/CoreML only. Comni seeing LLM / Vision / TTS on your RTX is expected.

This is a heavier workload than a normal LLM on the same GPU. Full Omni loads LLM + vision + audio + TTS + token2wav together. Q4_K_M is around ~9GB VRAM; on Windows a 12GB card also loses some memory to the desktop, so layers often get offloaded to CPU. That matches the slow first load, slow start → live, GPU at 13% or 0%, and a regular LLM feeling much faster.

“Answers once, then silence” is usually the live/duplex audio path getting stuck, not just low tok/s.

A new Comni app is in the works and should handle this Windows / 12GB setup more cleanly. Please try that build when it lands.

WhatABigWorld

5 days ago

•

edited 5 days ago

thanks for being so active and for working on solving hat

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment