Skip to main content

​
Synopsis

ollama pull MODEL

​
Description

The pull command downloads a model from a registry (default: Ollama’s public registry) to your local machine. Once pulled, the model can be used with ollama run and other commands. Models are downloaded in chunks with progress indication. If the download is interrupted, it will resume from where it left off.

​
Arguments

​
MODEL
string
required
Name of the model to pull. Can include:
  • Model name: llama3.2
  • Model with tag: llama3.2:8b
  • Model from custom registry: registry.example.com/mymodel:latest

​
Options

​
--insecure
boolean
default:"false"
Allow pulling from registries without HTTPS
ollama pull --insecure localhost:5000/mymodel

​
Model Naming

Models follow this naming convention: 
[REGISTRY/][NAMESPACE/]MODEL[:TAG]
  • REGISTRY: Optional, defaults to registry.ollama.ai
  • NAMESPACE: Optional, defaults to library
  • MODEL: Required model name
  • TAG: Optional, defaults to latest

​
Examples

ollama pull llama3.2              # Pulls library/llama3.2:latest
ollama pull llama3.2:8b           # Pulls library/llama3.2:8b
ollama pull mistral:7b-instruct   # Pulls library/mistral:7b-instruct

​
Examples

​
Pull Default Tag

Pull a model with the default latest tag: 
ollama pull llama3.2
Output: 
pulling manifest
pulling 8934d96d3f08: 100%
pulling 8c17c2ebb0ea: 100%
pulling 7c23fb36d801: 100%
pulling 2e0493f67d0c: 100%
pulling fa304d675061: 100%
pulling 42ba7f8a01dd: 100%
verifying sha256 digest
writing manifest
success

​
Pull Specific Tag

Pull a specific model variant: 
ollama pull llama3.2:8b

​
Pull from Custom Registry

Pull from a private or custom registry: 
ollama pull registry.company.com/custom-model:v1

​
Pull Without TLS

Pull from a local or insecure registry: 
ollama pull --insecure localhost:5000/test-model
Only use --insecure for trusted local registries. Never use it for public or untrusted sources.

​
Progress Output

During download, you’ll see progress for each layer: 
pulling manifest
pulling 8934d96d3f08: 45% ████████░░░░░░░░ 2.3 GB/5.0 GB
pulling 8c17c2ebb0ea: queued
Progress indicators show:
  • Layer digest (first 12 characters)
  • Percentage complete
  • Bytes downloaded / total size

​
Storage Location

Models are stored in:
  • Linux/macOS: ~/.ollama/models
  • Windows: %USERPROFILE%\.ollama\models
  • Custom: Set OLLAMA_MODELS environment variable

​
Model Layers

Ollama models consist of multiple layers:
  • Manifest: Model metadata and configuration
  • Model weights: Neural network parameters (largest layer)
  • Template: Prompt formatting template
  • Parameters: Model hyperparameters
  • System message: Default system prompt
  • License: Model license information

​
Resumable Downloads

If a download is interrupted (network issue, Ctrl+C, etc.), simply run ollama pull again. The download will resume from where it left off without re-downloading completed layers.

​
Disk Space

Before pulling, ensure you have sufficient disk space:
Model SizeApproximate Disk Space
1B params~1 GB
3B params~2 GB
7B params~4-5 GB
13B params~8-9 GB
34B params~20 GB
70B params~40 GB

​
Environment Variables

​
OLLAMA_HOST
string
default:"http://127.0.0.1:11434"
Ollama server address to send the pull request
​
OLLAMA_MODELS
string
default:"~/.ollama/models"
Directory where pulled models will be stored

​
Exit Codes

  • 0 - Success, model pulled completely
  • 1 - Error occurred (network issue, insufficient disk space, etc.)

​
Troubleshooting

​
Model Not Found

Error: model 'invalid-model' not found
Solution: Check the model name and tag. Browse available models at ollama.com/library

​
Insufficient Disk Space

Error: no space left on device
Solution: Free up disk space or change the models directory: 
OLLAMA_MODELS=/path/to/larger/disk ollama pull llama3.2

​
Network Timeout

Error: context deadline exceeded
Solution: Check your internet connection and try again. The download will resume automatically.

​
Registry Authentication

For private registries requiring authentication: 
# Sign in first
ollama signin

# Then pull
ollama pull registry.ollama.ai/private/model