OpenParable
Skip to content
← Back to skills

Minicpm5 Deploy Ollama

Apps & Automation
minicpm5-deploy-ollama

Minicpm5 Deploy Ollama is an Agent Skill for apps & automation published in the OpenBMB/MiniCPM repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Run MiniCPM5-1B via Ollama on macOS / Linux laptop using the released GGUF. Use when the user wants "ollama run", "ollama pull", a Modelfile-driven setup, or one-line laptop deployment.

View on GitHub ↗
QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

  1. Open your project in Codex.
  2. Copy and paste the prompt below.
  3. Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/OpenBMB/MiniCPM/blob/HEAD/skills/minicpm5-deploy-ollama/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/minicpm5-deploy-ollama, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

Deploy MiniCPM5-1B with Ollama

One-binary, no-Python laptop deployment. Consumes the released GGUF.

Required input

VarExampleDefault
GGUF_REPOopenbmb/MiniCPM5-1B-GGUFrequired
QUANTQ4_K_M (657 MB, recommended) / Q8_0 / F16Q4_K_M
MODEL_NAMEminicpm5-1bminicpm5-1b

Steps

1. Install Ollama (once)

brew install ollama                                # macOS
# or:
curl -fsSL https://ollama.com/install.sh | sh      # Linux

OLLAMA_FLASH_ATTENTION=1 OLLAMA_KV_CACHE_TYPE=q8_0 ollama serve &

2. Download the GGUF + write Modelfile

mkdir -p ~/${MODEL_NAME} && cd ~/${MODEL_NAME}

huggingface-cli download ${GGUF_REPO} MiniCPM5-1B-${QUANT}.gguf --local-dir .

cat > Modelfile <<EOF
FROM ./MiniCPM5-1B-${QUANT}.gguf

# MiniCPM5 chat template (matches release tokenizer)
TEMPLATE """{{- if .Messages -}}
{{- range .Messages -}}
<|im_start|>{{ .Role }}
{{ .Content }}<|im_end|>
{{ end -}}
<|im_start|>assistant
{{ end -}}"""

PARAMETER stop "<|im_end|>"
PARAMETER stop "</s>"

# Defaults tuned for nothink mode
PARAMETER temperature 0.7
PARAMETER top_p 0.95
PARAMETER num_ctx 8192
EOF

3. Create + run

ollama create ${MODEL_NAME} -f Modelfile
ollama run ${MODEL_NAME}

4. Validate via OpenAI-compatible API

curl http://localhost:11434/v1/chat/completions \
    -H "Content-Type: application/json" \
    -d '{
        "model": "minicpm5-1b",
        "messages": [{"role": "user", "content": "1+1=?"}],
        "temperature": 0.7, "top_p": 0.95, "max_tokens": 64
    }'

Expected: "2" in the reply.

Think mode

Default Modelfile is nothink. For think:

ollama run ${MODEL_NAME} --temperature 0.9 --top-p 0.95

Or bake it into a separate model tag by flipping temperature 0.7 to temperature 0.9 (top_p stays 0.95) and ollama create ${MODEL_NAME}-think -f Modelfile.think.

To force the auto-injected <think>\n prefix, use raw mode:

curl http://localhost:11434/api/generate -d '{
    "model": "minicpm5-1b",
    "raw": true,
    "prompt": "<|im_start|>user\n鸡兔同笼…<|im_end|>\n<|im_start|>assistant\n<think>\n",
    "options": {"temperature": 0.9, "top_p": 0.95}
}'

Common pitfalls

  • Error: invalid file magic: corrupted download. Re-run huggingface-cli download.

When NOT to use

  • Highest throughput on Mac → minicpm5-deploy-mlx (Q4 build)
  • GUI experience → minicpm5-deploy-lmstudio
  • NVIDIA GPU production → minicpm5-deploy-vllm

Reference

docs/deployment/ollama.md