Spaces:
Running
title: README
emoji: 🔷
colorFrom: indigo
colorTo: blue
sdk: static
pinned: false
Hologram Technologies
A web operating system that runs entirely in the browser, with no server behind it.
Hologram runs a full operating system, its apps, and its built-in assistant Q inside a single browser tab. Nothing is installed and no server is trusted: everything is delivered as content that verifies itself as it loads.
This organization publishes the language models that Q runs. Each model is split into content-addressed blocks (every file is named by the SHA-256 hash of its own bytes), so it can stream from any CDN and be checked, block by block, as it arrives. The models run in the browser on WebGPU, with no install and no inference server.
Models
| Model | Size | Precision | Best for |
|---|---|---|---|
| BitNet-2B-4T | 0.69 GB | 1.58-bit ternary | Fast general chat. Q's default. |
| Falcon-E-3B | 0.63 GB | 1.58-bit ternary | General chat with a little more depth. |
| TriLM-3.9B | 0.87 GB | 2-bit ternary | Best quality under a gigabyte. |
| Qwen2.5-Coder-7B | 3.4 GB | 3-bit | Writing code and using tools. |
| OLMoE-1B-7B | 3.6 GB | 4-bit, mixture of experts | Quick answers at a larger model's quality. |
| Dream-7B | 2.9 GB | 3-bit, diffusion | Experimental. Writes text in parallel passes. |
Every model is a re-encoding of a published open model; each repository names its source and license. To try one, open the live demo and pick a model from Q.