# Ollama: native apps and a local API

A maintainer starting point for native installation and local model serving. Use its linked documentation to check platform support, installed model identity and local/cloud behavior.

- Type: Deployment recipe
- Evidence: source_reported; site reproduced: false
- Model family: General
- Model: None
- Original author: ollama
- Original project: https://github.com/ollama/ollama
- Pinned document: https://github.com/ollama/ollama/tree/d1ed3fd13ac73f60cff00bbead8af24a68c4200d
- Source revision: d1ed3fd13ac73f60cff00bbead8af24a68c4200d
- Source commit time (not experiment date): 2026-10-17T01:45:46Z
- Checked: 2026-10-17
- Experiment date: None
- Runtime topics: Ollama
- Hardware topics: cpu, apple, nvidia
- Reported setup: {"system": null, "node_count": null, "parallelism": null, "kv_dtype": null, "context_limit_tokens": null, "experiment_date": null}
- Performance figures: attributed selections with per-point conditions; no cross-source ranking

## Conditions

Unknown is null. Recipe options are not measured benchmark settings.

```json
{
  "conditions": {
    "model_revision": null,
    "weight_hash": null,
    "runtime_execution_revision": null,
    "input_tokens": null,
    "output_tokens": null,
    "concurrency": null,
    "cache_state": null
  },
  "test_profiles": [],
  "recipe_parameters": null,
  "deployment_options": []
}
```

## Reading notes

- Select a platform-specific installation path.
- Capture the installed model digest rather than relying on a name alone.
- Check the client workflow separately from the local server.

## Limits

- Model, digest and actual hardware fit are not established by this reference.

## License declaration

{"declared_spdx": "MIT", "scope": "Repository declaration only; models, containers, datasets and per-file terms remain separate."}


## Editorial answer

A maintainer starting point for native installation and local model serving. Use its linked documentation to check platform support, installed model identity and local/cloud behavior.

By Run AI Here editorial · 2026-10-17

AI-assisted source reading and original editorial synthesis; no independent model execution.

## What is the smallest useful local check?

Install a recorded platform release, start the intended local app or server, and query its model list on the default loopback API port 11434. An empty model list is a valid response before weights are installed. Choose an already available model and keep its returned digest with the runtime version before sending one small public prompt.

Sources: https://github.com/ollama/ollama/blob/d1ed3fd13ac73f60cff00bbead8af24a68c4200d/README.md#L90

## Why does a model name need more provenance?

A tag can move while its friendly name stays the same. Capture the installed model digest, model origin and any imported conversion. The upstream README includes local model execution as well as integrations with agent applications; an integration name does not prove that the entire client workflow stays offline.

Sources: https://github.com/ollama/ollama/blob/d1ed3fd13ac73f60cff00bbead8af24a68c4200d/README.md#L78

## Which installation and client paths are separate?

The README provides distinct macOS, Windows, Linux and Docker routes, plus Python and JavaScript libraries. Select the documented path for the host rather than mixing shell commands across operating systems. Retain the previous installer and configuration so a version change can be reversed without deleting the model store.

Sources: https://github.com/ollama/ollama/blob/d1ed3fd13ac73f60cff00bbead8af24a68c4200d/README.md#L29

## How should local-only behavior be verified?

A loopback endpoint establishes where that HTTP request goes. Cloud features, a selected cloud model or remote functions in the client are separate boundaries. Use the local-only guide to disable the documented server cloud feature and test the exact already-downloaded workflow offline. This reference has not inspected the reader’s machine or run its models.

Sources: https://github.com/ollama/ollama/blob/d1ed3fd13ac73f60cff00bbead8af24a68c4200d/README.md#L70