Elephas Inbuilt Offline Models for 100% Privacy
Using Elephas Inbuilt Offline Models
Using Elephas Inbuilt Offline Models
Elephas now comes with inbuilt offline AI models, giving you the ability to use Elephas without an internet connection. With these models, you can run chat and indexing locally on your machine, ensuring complete privacy and faster performance.
You need Apple Silicon (M1 and later) to use this feature.
Creating a Brain with Offline Models
The primary way to leverage Elephas offline models is through Brain creation. Here's how to set up a Brain using offline models:
- Open the Create Brain window.
- Under Mode, select Offline.
- Choose your models:
- Indexing Model → Select from offline embedding models (e.g., multilingual-e5-large). multilingual-e5-large offers better quality results than small variant.
- Default Chat Model → Select from offline chat models (e.g., llama3.2:3b). llama3.2:3b offer good quality results. Uses around 2 GB of memory when running.
- Click Create Brain to finish.
When you create a Brain with an offline model, Elephas takes care of the heavy lifting for you. If the model isn’t already on your machine, it will automatically download and start in the background — no extra clicks needed. Just set it and go!
File Indexing in Offline Brains
Once your offline Brain is created, the next step is to add and index your files.
Here’s an example of indexing Apple Notes into your Brain:
- Open the Brain you just created.
- Navigate to the Integrations section and select Apple Notes.
- Choose the Apple Notes folders or individual notes you’d like to include, then click Connect Notes.
- Elephas will automatically index your files locally using the selected offline embedding model.
Your private Apple Notes will never leave the system. Everything is processed locally—offline and secure.
For a step-by-step walkthrough, refer to Integrate Apple Notes with Elephas Super Brain
Using Super Chat with Offline Brains
After your files are indexed, you can start chatting with your Brain — fully offline.
- Open your offline Brain.
- Switch to the Super Chat tab.
- Ask questions based on your indexed files, and Elephas will respond using the offline chat model.
For detailed instructions, refer to our Elephas Super Chat support documentation.
Memory Management with Offline Models
The following model management guide is optional, as Elephas handles the model download, load automatically. This is only needed if you want to stop the model to save memory.
Elephas Offline by default runs only one indexing model and only one chat model at any point of time. This is done to save memory.
If you are using an Elephas built-in offline model as your chat model, keep in mind that:
- These models use a significant amount of memory.
- Once your chat session is complete, you can turn off or stop the offline chat model from Elephas Preferences → Offline tab → Chat Models.
This helps free up memory and ensures smoother performance for other tasks.
Managing Offline Models
Accessing Offline Models in Elephas
- Open Elephas Preferences.
- Navigate to the Offline tab.
- You’ll see two sections:
- Chat Models – for conversations and general AI tasks.
- Indexing Models – for embeddings, semantic search, and retrieval.
Downloading a Model
Some models are not pre-installed. To use them:
- Click Download next to the model.
- Wait for the download to finish.
- Once ready, click Start to activate it.
Starting a Model
- Find the model you want to use (e.g., Multilingual E5 Large).
- Click Start.
- The model will load locally and show as Running.
🔄 Only one offline model can be active at a time. Starting a new model will automatically stop the currently running one.
Stopping a Model
If a model is currently running, you’ll see a red Stop button on its card.
- Click Stop to deactivate the model.
- The status will change back to Ready to start.
- You can then start another model if needed.
Only one offline model can run at a time. Stopping the current model is required before starting a new one.
Deleting a Model
If you no longer need a downloaded model:
- Click the 🗑 Trash icon on the model card.
- Confirm the deletion.
- You can re-download it anytime if needed.
Configuring in Model Settings
Once you’ve started an offline model, you can choose it in Elephas Model settings:
- Open Preferences → Model Settings
- Select the model under Elephas Offline provider
- Pick the model you want for:
- Chat → choose from offline chat models (e.g., Llama 3.2 3B Instruct).
- Indexing → choose from offline embedding models (e.g., Multilingual E5 Large).
With inbuilt offline models, Elephas gives you the power of AI anytime, anywhere — completely private and under your control.