Run Llama 3.2 3B in your browser
Meta’s 3B instruct model. A dependable all-rounder for laptops with 8 GB or more.
Use Llama 3.2 3B in the browser All models Downloads once · then works offline
- Family
- Llama
- Published by
- Meta
- Parameters
- 3B
- Quantisation
- q4f16_1 · 4-bit
- Download
- 1.7 GB · 58 files
- GPU memory
- 2.2 GB
- Context window
- 4k tokens
- Reasoning
- —
- Licence
- Llama 3.2 Community License
- Base model
- meta-llama/Llama-3.2-3B-Instruct
- Good at
- Chat · Multilingual
How big is the Llama 3.2 3B download?
1.7 GB for the q4f16_1 build, in 58 files. It downloads once and stays in your browser cache, so every later visit starts straight away and works offline.
What does my device need to run Llama 3.2 3B?
A browser with WebGPU (a recent Chrome, Edge, Firefox or Safari) and about 2.2 GB of GPU memory. Before anything downloads, the tool checks your GPU, memory and free space and tells you whether Llama 3.2 3B fits.
Is anything I type sent to a server?
No. Llama 3.2 3B runs on your own GPU inside the browser tab. The only network traffic is the one-time download of the model files from Hugging Face; your prompts and the answers never leave the device, and the chats are stored in this browser.
Which build of Llama 3.2 3B should I choose?
Start with q4f16_1. It is the smallest download and the fastest to run. The f32 builds are larger and use more memory, but they run on GPUs without 16-bit shader support. Builds marked 1k hold a shorter conversation in exchange for less memory.
How long a conversation can Llama 3.2 3B hold?
Its context window is 4k tokens. Longer conversations still work: once the conversation no longer fits, the oldest turns are dropped and you see a notice.
What licence is Llama 3.2 3B released under?
The weights are Llama 3.2 Community License, taken from the model card of meta-llama/Llama-3.2-3B-Instruct. Check the licence yourself before using it commercially.
Figures read from the WebLLM 0.2.84 catalog and the model card on 2026-09-03.
