Question 1

What is LM Playground?

Accepted Answer

LM Playground lets you run large language models directly on your Android device. All processing happens locally – no cloud servers needed.

Question 2

What models are supported?

Accepted Answer

The app supports models in GGUF format with Q4_K_M quantization, including Qwen 3, Gemma 3, Llama 3.2, Phi-4 mini, and DeepSeek R1 Distill in various sizes.

Question 3

How much storage space do I need?

Accepted Answer

Model sizes vary from about 500 MB (small models like Qwen 3 0.6B) to several GB (larger models like DeepSeek R1 7B). Make sure you have enough free space before downloading.

Question 4

Does the app require an internet connection?

Accepted Answer

Internet is needed to download models, and — if you turn them on — for the optional web tools (web search and fetch page). Apart from that, once a model is downloaded it runs completely offline on your device.

Question 5

Is my data private?

Accepted Answer

Your conversations are processed locally on your device and are never sent to a server by the app. The one exception is the optional web tools: if you enable web search or fetch page, the model's search queries and the pages it opens are sent to those websites (search uses DuckDuckGo). Tools are off by default, so nothing leaves your device unless you turn them on. See the Privacy Policy for details.

Question 6

Why is model loading slow?

Accepted Answer

Larger models take more time to load into memory. Loading times depend on your device's hardware. Once loaded, the model stays in memory until you unload it.

Question 7

Which devices work best?

Accepted Answer

Devices with more RAM can run larger models. For best performance, use a device with at least 6 GB of RAM for small models and 8+ GB for larger ones.

Question 8

Can I load a custom GGUF model?

Accepted Answer

Yes. Place your .gguf file in the storage folder selected in Settings → Models (the same folder used for downloads). The app will pick it up automatically and show it in the model selector alongside the built-in catalog. Chat template and tokenizer settings are read from the GGUF metadata. If a specific model doesn't work, please open an issue on GitHub.

Question 9

Can I change where models are stored?

Accepted Answer

Yes. Go to Settings, then Models, and use the “Change Folder” option to select a different storage location.

Question 10

How do I delete a model?

Accepted Answer

Go to Settings, then Models. In the “Downloaded” section, tap the delete icon next to the model you want to remove.

Question 11

What are system prompts?

Accepted Answer

A system prompt is a reusable instruction that shapes how the model responds — for example, “Always answer in French” or “Be concise.” Create and manage them in Settings → System Prompts, then apply one to your chat. They are stored on your device and sent only to the local model.

Question 12

What are tools and which ones are available?

Accepted Answer

Tools let capable models take actions during a reply. Three are available: Run JavaScript (math, dates, and text processing in an on-device sandbox), Web search (finds current information on the web), and Fetch web page (reads a page so the model can summarize or quote it). Turn them on in Settings → Tools. Each is off by default and works only with models that support tool calling — look for the hammer badge in the model selector.

Question 13

Do the web tools send my data online?

Accepted Answer

Only when you enable them. Web search sends the model's query to DuckDuckGo, and Fetch web page downloads the pages the model chooses to open — those requests go to the websites themselves, not to the developer. Run JavaScript stays entirely on your device with no internet access. With every tool off (the default), nothing leaves your device. See the Privacy Policy for details.

Question 14

Will the app notify me when a response is ready?

Accepted Answer

Yes. While the app is in the background, a notification tracks the model's status — loaded, generating, then response ready — and lets you copy or share the reply. You can also enable a short completion chime in Settings → Sound and Haptic. Model downloads show their own progress notification with speed and ETA.

Run large language models
on your phone.

Private by construction. Fast by design.

Fully offline

Your roles, saved

Tune every knob

Save anywhere

Pick a brain. Tap download. Chat.

Three taps from app open to first token.

Pick a model

Download and resume

Chat locally

Set the role once.
Reuse it everywhere.

Let capable models
take real actions.

Run JavaScript

Web search

Fetch web page

Speaks your language.
Right out of the box.

What you actually
wanted to know.

Any model in your pocket.

Tweaks