How to create a Modelfile to customize model parameters
Ollama installed, base model already pulled to local library
What this does
Creates a named model variant with customized inference parameters, system prompts, and template overrides using a declarative Modelfile. After this guide a tailored version of the base model will be available for specialized use cases.
Steps
Create a Modelfile in a working directory. The FROM directive specifies which base model to extend.
cat > Modelfile << 'EOF' FROM llama3.2 PARAMETER temperature 0.3 PARAMETER num_ctx 8192 PARAMETER top_p 0.9 SYSTEM """You are a coding assistant. Always include code examples.""" EOFLow temperature (0.3) produces deterministic output; higher values increase creativity.
Create the custom model variant. Registers the customized model under a new name.
ollama create coding-llama -f ModelfileExpected output:
transferring model datafollowed bysuccess.Run the custom model and verify behavior. Launches the variant with the baked-in parameters.
ollama run coding-llama "Write a hello world in Python"Expected output: A response that includes a code example.
- Record the local run evidence. Save the exact command, runtime or package version, model name if applicable, and observed output so the result can be reproduced later.
Verification
ollama show coding-llama
# Expected: output includes PARAMETER block with temperature 0.3 and num_ctx 8192
Common failures
unable to locate model- FROM references a model not in the local library; runollama pull <name>first.syntax error in Modelfile- Missing newline after a directive or unbalanced quotes; check the file character by character.system prompt ignored- Some model architectures require a TEMPLATE directive before SYSTEM will take effect.duplicate model name- A model with the target name already exists; choose a different name or delete the existing one.