The Model Hub

MyLLM's Model Hub is your gateway to different AI personalities and capabilities. With 6 model families and 10+ models available, you can choose the perfect AI for any task — from quick answers to deep reasoning.

Downloading Models

From the Built-in Hub

The easiest way to get started:

Open MyLLM and go to the Model Hub tab
Browse available models with descriptions and specs
Tap "Download" on any model
The download runs in the background — you can keep using the app
Once complete, tap to activate the model

Downloads are handled by Android's WorkManager, which means:

Background downloads — Continue using your phone while downloading
Resume support — If your connection drops, the download resumes automatically
Notification progress — Track download progress from your notification bar

Custom GGUF Models

Already have GGUF models from HuggingFace or other sources? MyLLM supports importing custom models:

Download any GGUF model file to your device
In MyLLM, go to Model Hub > Import
Select the GGUF file from your storage
The model appears in your library, ready to use

This means you can use any model that's been converted to GGUF format — not just the ones we list.

Supported Model Families

Qwen 3.5 (Recommended)

Alibaba's latest generation. Excellent multilingual support including Hindi, Chinese, Arabic, and more. Best overall quality-to-size ratio.

0.6B — Lightning fast, basic quality. Great for quick answers.
1.5B — Good balance for mid-range phones. Handles most tasks well.
4B — Our top pick. Great quality at reasonable speed.
8B — Premium quality. Needs 6+ GB RAM but delivers impressive results.

Llama 3.2

Meta's open-source champion. Strong English performance and creative capabilities.

1B — Fast and reliable for English conversation
3B — Excellent quality, particularly for creative and narrative tasks

Gemma 2

Google's compact powerhouse. Trained on high-quality data with strong factual accuracy.

2B — Punches above its weight. Excellent for factual Q&A.

Phi-3.5 Mini

Microsoft's reasoning specialist. Outstanding code generation and logical thinking.

3.8B — Best-in-class for code among models this size.

SmolLM2

HuggingFace's efficiency champion. Designed to run well on resource-constrained devices.

1.7B — Reliable everyday assistant. Great on budget phones.

DeepSeek R1

The reasoning expert. Uses chain-of-thought internally for better problem-solving.

1.5B — Specializes in step-by-step reasoning and complex analysis.

Switching Between Models

One of MyLLM's best features is instant model switching:

Tap the model name in the chat header
Select a different downloaded model
The new model loads in seconds
Continue your conversation or start a new one

Why Switch Models?

Quick question? Use a small, fast model (0.6B-1.5B)
Complex coding task? Switch to Phi-3.5 or Qwen 4B
Creative writing? Llama 3.2 3B excels at narrative
Need reasoning? DeepSeek R1 for step-by-step analysis
Low battery? Smaller models use less power

Storage Management

Models can be large, so MyLLM includes storage management:

See model sizes before downloading
Delete unused models to free space
Storage usage overview in settings
Cache cleanup for temporary inference files

The GGUF Format

All models use the GGUF format, which stands for GPT-Generated Unified Format. It's the industry standard for running LLMs on local hardware, offering:

Efficient memory mapping — Models load faster and use less RAM
Quantization support — Multiple precision levels in a single format
Metadata included — The file contains all model information
Cross-platform — Same file works on any device with llama.cpp

Explore the Model Hub →

Model Hub: Download, Manage, and Switch AI Models Instantly