Ollama Model Library Scraper

Ollama Model Library Scraper

An easy-to-use tool designed to collect data from the open model catalog on Ollama. It helps you search models, grab details, check available tags, and inspect file sizes. You can easily export your results into spreadsheets like JSON, CSV, or Excel.

Whether you are looking up a single model or compiling a list, this tool processes requests smoothly, pauses between tasks to avoid overcrowding servers, and logs errors cleanly if a specific request fails so your whole run keeps going.

Use cases

• Model catalog: Build and maintain a searchable dataset of available models, descriptions, and popularity metrics.
• Local setup: Verify parameter sizes, available tags, and context windows before pulling models to your local environment.
• Capability filtering: Quickly discover models that support specific features like tools, vision, embeddings, or reasoning.
• Spreadsheet reporting: Export clean datasets directly to CSV or Excel for analysis or team sharing.
• Catalog tracking: Run periodic checks to monitor updates, new tags, or shifts in model popularity.

Input options

Parameter	Type	Default	Description
mode	string	search	Choose between search, model_details, or model_tags.
query	string	llama	Search text to filter models.
models	array	llama3.2	List of specific model names for details or tags mode.
capability	string	empty	Optional filter for features like tools, thinking, vision, or embedding.
order	string	popular	Sort order options: popular, newest, or name.
maxItems	integer	100	Target number of items to collect. Max limit is 1000.
requestDelaySecs	number	1	Delay between requests in seconds to keep execution smooth.
timeoutSecs	integer	300	Overall run timeout in seconds.
requestTimeoutSecs	integer	30	Timeout per individual request in seconds.
userAgent	string	B2BLeadGenerationResearch/1.0	Custom contact string sent with requests.
proxyConfiguration	object	Off	Optional proxy settings if required for your setup.

Example input

{
  "mode": "search",
  "query": "llama",
  "order": "popular",
  "maxItems": 100,
  "requestDelaySecs": 1,
  "timeoutSecs": 300,
  "requestTimeoutSecs": 30,
  "proxyConfiguration": { "useApifyProxy": false }
}

Collected data fields

The tool gathers key catalog attributes and formats them into a clean structured output:

{
  "name": "llama3.1",
  "url": "[https://ollama.com/library/llama3.1](https://ollama.com/library/llama3.1)",
  "modelSummary": "Llama 3.1 is a state-of-the-art model from Meta available in 8B, 70B and 405B parameter sizes.",
  "capabilities": ["tools"],
  "sizes": ["8b", "70b", "405b"],
  "downloads": 120225507,
  "downloads_display": "120.2M",
  "source_url": "[https://ollama.com/search?q=llama](https://ollama.com/search?q=llama)",
  "scraped_at": "2026-10-12T00:00:00+00:00"
}

When running in tags mode, additional detail fields are populated: tag, digest, size, context_window, input, and updated.

If an error occurs while fetching a specific item, an error row is recorded without halting the execution:

{
  "error": "Unknown model id",
  "url": "[https://ollama.com/library/missing-model](https://ollama.com/library/missing-model)",
  "status": 404,
  "message": "Unknown model id",
  "source_url": "[https://ollama.com/library/missing-model](https://ollama.com/library/missing-model)",
  "scraped_at": "2026-10-12T00:00:00+00:00"
}

Field	Type	Description
name	string	Model name identifier.
url	string	Web link to the model or tag.
modelSummary	string	Summary description of the model.
capabilities	array	Badges indicating supported capabilities (e.g., tools, vision).
sizes	array	Available parameter sizes (e.g., 8b, 70b).
downloads	integer	Numeric download count.
downloads_display	string	Formatted download count (e.g., 120.2M).
tag	string	Tag name (e.g., llama3.2:latest).
digest	string	Short hash digest.
size	string	Disk file size label (e.g., 2.0GB).
context_window	string	Maximum context length label (e.g., 128K).
input	string	Input type supported (e.g., Text).
updated	string	Time since last update.
source_url	string	Origin URL processed for the record.
scraped_at	string	Timestamp of collection in UTC.
error	string	Error label (only present on failed items).
status	integer	HTTP status code on error records.
message	string	Error detail explanation.

How it works

1. Reads your input configuration and initializes the workload.
2. Iterates through search results page by page (around 20 items per page), maintaining selected filter parameters.
3. Pauses briefly between actions according to requestDelaySecs and retries transient failures.
4. Removes duplicates by model name or tag and outputs data in structured batches until reaching maxItems or time limits.

Performance considerations

This actor is focused strictly on catalog information and text summaries. It does not download model binary files or weights, ensuring fast and lightweight operations.

FAQ

Can I export the collected data directly to spreadsheet formats? Yes, results can be downloaded in JSON, CSV, or Excel formats.

What happens if a requested model isn't found? The scraper logs a single error record containing the URL and error status, then seamlessly proceeds to process the remaining models.

How many records can be collected in a single run? The default limit is set to 100, with a configurable cap up to 1000 items per run.

Is a proxy necessary? Proxies are disabled by default and are generally not required unless you encounter IP restrictions.

Trademark Notice

Ollama Model Library and related product names, logos, and trademarks belong to their respective owners. Their inclusion in this document is purely for descriptive purposes to identify the source dataset and does not denote endorsement or official partnership.

Disclaimer

This Actor is an independent collection tool and is not affiliated with, endorsed by, or sponsored by Ollama Model Library or its parent entities. Users are responsible for ensuring their usage aligns with relevant terms of service, rate guidelines, and licensing terms.

Is web scraping legal?

This Actor gathers publicly accessible data from the catalog without bypassing security mechanisms or restrictions. Gathering open data is generally permissible, but you should take standard precautions depending on your use case:

• Personal data: If your collected datasets ever involve personal details, ensure compliance with relevant privacy regulations like GDPR or CCPA.
• Copyright: Always treat underlying descriptions and text respectfully and avoid re-publishing full copyrighted materials inappropriately.
• Terms of use: Ensure your request frequency stays within reasonable limits and complies with standard internet guidelines.
• Legal advice: This overview provides general context and does not constitute formal legal counsel.

To learn more, read Apify's guide: Is web scraping legal?