Managing Models
Overview
Ollama lets you save your work and reuse it later.
When you save a session, Ollama creates a new model that includes your current settings, system prompt, and chat history.
- Save a session
- Load a saved session
- Reuse prompts and settings
- Manage local models
- Remove unused models
This makes it easy to continue previous work without setting everything up again.
Save a Session
You can save the current session using the /save command.
Under the hood, Ollama actually saves a copy of the original model with your custom settings and history.
In the example below, I add my name "Eden" into the conversation before saving.
Hi, I am Eden.
Save the session:
## "s1" can by any name you choose for the saved model.
/save s1
Output:
Created a new model 's1'
The saved model now contains the current chat history and settings.
Now you can exit the session or start a new one without losing your work.
/bye
Load a Saved Session
Start the model again:
ollama run <your_model_name>
Load the saved session.
/load s1
Now ask a question:
What's my name?
Output:
Your name is Eden.
Even though the name was not entered in the current session, Ollama remembers it because it was saved previously.
This makes saved sessions useful for continuing previous conversations.
What Gets Saved
When you save a session, Ollama stores more than just the chat history.
- Chat history
- System prompt
- Parameter settings
- Session configuration
The saved model starts with all of these settings already loaded.
You can think of it as a customized version of the original model.
Save a System Prompt
System prompts can also be saved.
In the example below, a system prompt is configured before saving.
/set system You are a friendly customer support agent. Do not approve refunds.
Save the session:
/save s1
Exit the session:
/bye
Start a new session and load the saved model:
ollama run <your_model_name>
/load s1
Verify the system prompt:
/show system
Output:
You are a friendly customer support agent. Do not approve refunds.
The system prompt is restored because it was saved with the model.
View Available Models
Ollama stores downloaded and custom models locally.
To see all available models, run:
ollama list
Output:
NAME ID SIZE MODIFIED
s1:latest df44b2559dee 9.6 GB About a minute ago
gemma3:4b a2af6cc3eb7f 3.3 GB 28 minutes ago
gemma4:e4b c6eb396dbd59 9.6 GB About an hour ago
minimax-m3:cloud d03a959f45c0 - About an hour ago
The list includes:
- Downloaded models
- Custom saved models
- Derived models
- Current session models
This is the main command used to see what models are available.
View Running Models
Ollama keeps recently used models in memory.
To see currently running models:
ollama ps
Output:
NAME ID SIZE PROCESSOR CONTEXT UNTIL
gemma4:e4b c6eb396dbd59 9.5 GB 100% CPU 4096 3 minutes from now
Notice that it has the "UNTIL" column which shows how long the model will remain in memory before it is automatically removed.
Run a Saved Model Directly
You can run a saved model without loading it manually.
In the example below, s1 is a saved model.
ollama run s1
Output:
>>> Hi, I am Eden.
Hi Eden! It's nice to meet you. 😊
How can I help you today? Is there anything specific you're looking to chat about or work on?
>>> What's my name?
Your name is Eden. 😊
>>>
The model starts with all saved settings and history already loaded.
This is often faster than loading a session manually.
Inspect a Model
You can view information about a model without starting it.
In the example below, gemma4:e4b is inspected.
ollama show gemma4:e4b
Output:
Model
architecture: gemma3
parameters: 4B
quantization: Q4_K_M
This is useful when you want to learn more about a model before running it.
Inspect a Custom Model
Custom models can also be inspected.
In the example below, s1 is a saved model.
ollama show s1
Output:
Model
architecture: gemma4
parameters: 4B
This helps identify:
- Which base model was used
- Model size
- Model configuration
It is useful when you forget what a custom model was created from.
Remove a Model
Unused models can be removed from your system.
In the example below, s1 is deleted.
ollama rm s1
Output:
Deleted 's1'
Use this command to clean up old or unused models.
Storage Usage
Saved models do not always create a full copy of the original model.
Ollama stores only the differences between the original model and the saved version.
For example:
- Original model:
gemma4:e4b - Saved model:
s1
Ollama reuses the original model files and only stores the additional settings and chat history.
This helps reduce disk usage while still allowing multiple custom versions of the same model.