Privacy
Keep your AI workflows closer to your own machine.
Checking latest release…
Local AI, Built for Control and Performance.
Run powerful language models on your machine with a polished desktop experience, live performance monitoring, flexible model support, and private local workflows.
View on GitHub
LOCAL-FIRST
PRIVATE
GGUF · BIN · GGML
DOCUMENT CONTEXT
WINDOWS DESKTOP
LLAMA.CPP
Why local
AI that stays closer to you.
Cloud AI is convenient, but local AI gives you more direct control over your data, your models, your runtime, and how your hardware is used.
Control
Choose the model, runtime, and configuration that fits your workflow.
Performance
See how your hardware is performing while the model runs.
Flexibility
Experiment with compatible local models without being locked into one provider.
Designed for local-first workflows.
Features
Everything you need for local AI.
01
AI Chat
An intuitive conversational workspace designed for local models.
02
Model Management
Work with the models you actually want to run.
03
Live Performance
Know what your machine is doing.
04
Local sessions
Organize, rename, pin, and remove conversations in the app.
05
Configuration
Control how your local AI environment behaves.
06
Desktop Experience
A polished desktop application built for Windows.
07
Model Library
Find and download compatible models directly in OffyAI through its connection to the public Hugging Face API.
Product tour
See OffyAI from the inside.
Ready. Model loaded and listening on the local runtime.
Explain what changed in this diff.
The file context is attached to the conversation and the local model is ready to respond.
Message OffyAI…
offyai.ggufActive
Add compatible local modelsOptional
Local runtime readyReady
CPU42%
Memory6.8 GB
GPU37%
Tokens/s28.4
General Q&A Today
Refactor plan Yesterday
Local model notes 3 days ago
Draft outline Last week
Local server URL
http://localhost:8080Theme
DarkContext size
4096Streaming
EnabledPerformance monitoring
Don't just run a model. Understand it.
OffyAI exposes useful runtime information while AI workloads are running, so you can see how your hardware responds in real time.
CPU
42%
Memory
6.8 GB
GPU
37%
Generation
28.4 tok/s
Response
1.8s
Metrics shown above are illustrative UI examples, not live statistics from this website.
Model support
Built-in model support.
OffyAI includes a protected default model named offyai.gguf and lets you find and download compatible models through its library connected to the public Hugging Face API.
GGUF
Local runtime
Default model
- Built-in offyai.gguf model
- Local inference workflow
- Model upload and selection
- Model library connected to the public Hugging Face API
- Download compatible models in OffyAI
Load
Chat
Monitor
How it works
From install to inference.
-
01
Choose or download a model
Select a compatible model from the library or place one in the local models directory.
-
02
Start the local runtime
OffyAI connects to the configured llama.cpp-compatible local backend.
-
03
Start chatting
Choose the model and interact through the desktop interface.
-
04
Monitor & manage
Track performance, manage conversations, and configure the environment.
Privacy
Your machine. Your models. Your control.
OffyAI is designed to keep local inference and app data on your machine, subject to the model, runtime, and external services you choose to configure.
Local model execution
Local conversation sessions
Local configuration
No mandatory cloud inference
User-controlled model files
Optional external integrations
API integrations, updates, downloads, or other configured external services may involve network communication.
Architecture
Under the hood.
OffyAI application stack
Electron
OffyAI Desktop UI
Local Application Logic
llama.cpp-compatible Runtime
Local Model
User Hardware
System requirements
What you'll need.
Operating System
Windows 10 / 11
Storage
Enough space for the application and any selected models
Memory
Depends on model size and quantization
GPU
Optional, depending on runtime configuration
CPU
Modern 64-bit processor recommended
Model
A compatible local GGUF, BIN, or GGML model
Large models require substantially more memory and compute resources.
Download
Run OffyAI on your Windows machine.
Checking for the latest release…
- Latest version
- —
- Release date
- —
- Installer name
- —
- Installer size
- —
What's new
Latest release notes.
—
—
Release information unavailable
Release notes will appear here once fetched from GitHub.
Installation
Get up and running.
- Download the Windows installer.
- Run the
.exefile. - Install OffyAI.
- Place compatible model files in the models directory.
- Configure or select the model.
- Start chatting.
- Monitor performance.
Security note: if Windows displays a security warning, verify that you downloaded the installer from the official GitHub release before proceeding.
FAQ
Common questions.
OffyAI is a Windows desktop application that provides a polished, ChatGPT-like interface for running compatible language models on your own machine, alongside model management and performance monitoring tools.
Yes. OffyAI is designed for local-first workflows, connecting to a llama.cpp-compatible runtime running on your machine.
Core chat with a local model does not require a mandatory cloud dependency. Downloads, updates, and any external services you choose to configure may still require network access.
No. OffyAI provides a ChatGPT-like interface but runs compatible local language models through a llama.cpp-compatible runtime rather than depending on ChatGPT itself.
GGUF, BIN, and GGML, subject to llama.cpp compatibility.
Yes. OffyAI includes a model library connected to the public Hugging Face API, so you can find and download compatible models for local use.
GGUF is a model file format commonly used with llama.cpp-based local inference runtimes.
Yes. OffyAI ships with a built-in default local model file named offyai.gguf, which is the initial model used when the app starts.
Conversation sessions are managed in the desktop app, with storage behavior determined by the app and runtime environment.
Application settings, including theme, server URL, and model configuration, are stored locally.
Yes. Place a compatible GGUF, BIN, or GGML model in the local models directory and select it in the application.
GPU usage depends on your runtime configuration and hardware; GPU monitoring is shown when available.
Requirements depend on the model you choose. See the System Requirements section for general guidance.
Check the project's GitHub repository or release information for current licensing terms.
Report bugs through the GitHub Issues page for this repository.
Feature requests can also be submitted through GitHub Issues.
Use the Download section on this page, or visit the GitHub Releases page directly.
Help
Troubleshooting.
Model not found
Verify the model path and confirm the file is a supported format (GGUF, BIN, or GGML).
Model does not appear
OffyAI validates available local model files; an invalid or unreadable file may not be listed.
Slow generation
Speed depends on model size, quantization, CPU, GPU, RAM, and runtime configuration.
High memory usage
Larger models require more memory; consider a smaller or more quantized model.
Local server not connecting
Verify the server URL, confirm the runtime is running, and check the port configuration.
Application does not start
Try reinstalling, verify you downloaded from the official source, check Windows compatibility, and review release notes or open issues.
Open source
Built in the open.
—Stars
—Forks
—Open issues
Take AI back to your machine.
Explore local models, monitor performance, and build an AI workflow around your own hardware.