ollm models

ollm models covers discovery, inspection, local materialization, and path reporting.

Subcommands

• list
• info
• download
• path

list

Discovery view over:

• built-in aliases
• local materialized models

Examples:

ollm models list
ollm models list --installed

info

Inspect a single model reference and its runtime-plan truth:

ollm models info llama3-1B-chat --json
ollm models info Qwen/Qwen2.5-7B-Instruct --json

download

Materialize a downloadable model reference locally. The downloader stores only the runtime-critical artifacts that ollm needs to inspect, plan, and execute the model locally rather than a full Hugging Face repository snapshot. For sharded repos, the managed materialization is only considered complete when every shard referenced by the safetensor index is present. Gated Hugging Face repos still require accepted terms and a token with weight access:

ollm models download llama3-3B-chat
ollm models download Qwen/Qwen2.5-7B-Instruct

path

Print the local materialization path for a reference that resolves to local storage.