AlphaChat — Run the Impossible. Privately.

Downloads

AlphaChat Desktop App

The desktop chat interface — chat, image analysis, workflows, and model management.

Platform	Size	Link
Windows x64	13 MB	Download

AlphaLlama AI Engine

The local AI inference engine — runs models on your GPU. Required for local inference. Optional if using cloud-only mode.

Platform	Size	Link
Windows x64	87 MB	Download
Linux x86_64	84 MB	Download

1. Install the Engine

Linux (x86_64)

# Download
curl -L https://alphachat.com/downloads/alphallama-linux-amd64.tar.gz -o alphallama.tar.gz

# Extract and set library path
tar xzf alphallama.tar.gz
export LD_LIBRARY_PATH=$(pwd)

Extracts to current directory. No root required. CUDA runtime bundled.

Supports all NVIDIA GPUs (Turing → Blackwell). Falls back to CPU if no GPU found.

Windows (x64)

# Download
curl -L https://alphachat.com/downloads/alphallama-windows-x64.tar.gz -o alphallama.tar.gz

# Extract
tar xzf alphallama.tar.gz

# Run
alphallama.exe -m model.gguf --port 18080 -ngl 999

Extracts to current directory. Includes CUDA runtime (ggml-cuda.dll).

Supports all NVIDIA GPUs (Turing → Blackwell). Falls back to CPU if no GPU found.

macOS

Coming soon. Join the waitlist to get notified.

2. Register (free)

AlphaLlama requires a free account. Register at alphachat.com/auth/register, then save your license token:

# After registering on alphachat.com, save your license token:
mkdir -p ~/.alphachat
echo "YOUR_LICENSE_TOKEN" > ~/.alphachat/license.jwt

Free tier: 2M token context, unlimited queries. No credit card needed.

3. Download a Model

# Download a model from HuggingFace (example: Qwen3.6 35B MoE)
# Any GGUF model works — see Models page for full list
curl -L https://huggingface.co/unsloth/Qwen3.5-35B-A3B-GGUF/resolve/main/Qwen3.5-35B-A3B-Q4_K_M.gguf -o model.gguf

Browse all supported models on the Models page. Any GGUF file works.

4. Run the Server

# Start AlphaLlama (auto-detects GPU)
./alphallama -m model.gguf --port 18080 -ngl 999

# Open the web UI in your browser
# http://localhost:18080

The server starts a web UI at the specified port. -ngl 999 offloads all layers to GPU.

5. Query via API

# OpenAI-compatible endpoint
curl http://localhost:18080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen","messages":[{"role":"user","content":"hello"}]}'

# Ollama-compatible endpoints also work:
# /api/chat, /api/tags, /api/ps, /api/show

Compatible with any app that uses the OpenAI SDK — just change the base URL to http://localhost:18080.

6. Advanced Options

# Set context size (default: 4096)
./alphallama -m model.gguf --port 18080 -ngl 999 --ctx-size 32768

# Run multiple parallel requests
./alphallama -m model.gguf --port 18080 -ngl 999 --ctx-size 32768 --parallel 4

# Bind to all interfaces (for remote access)
./alphallama -m model.gguf --host 0.0.0.0 --port 18080 -ngl 999

# Skip license check (cloud/RunPod deployment)
./alphallama -m model.gguf --port 18080 -ngl 999 --no-license

6. Pricing

All plans run on your GPU. Unlimited queries. Data never leaves your device.

Plan	Price	Context
Free	$0	2M tokens
Pro	$19/mo	50M tokens
Business	$0.14/$0.28 per MTok	Unlimited
Manufacturer	$0.105/$0.21 per MTok	Unlimited (25% off)

Full pricing details →

Fully private. When running locally, your data never leaves your device. No cloud, no network, no third parties. Just your GPU.

Quick Start