DS4: Complete Guide to Running DeepSeek 4 Flash and PRO Locally
DS4 is a local inference engine for DeepSeek 4 Flash and PRO by antirez — the Redis creator. It runs on Metal, CUDA and ROCm for fully private, offline LLM inference on your own hardware.
🚀 Want to deploy DS4 yourself?
Docker configs, system requirements, and installation guides — all on one page.
View DS4 Tool Page →What Is DS4?
DS4 is a local inference engine for the DeepSeek 4 Flash and PRO model family, created by antirez — the same developer who built Redis. It brings state-of-the-art LLM inference to your own hardware, with first-class support for Apple Metal, NVIDIA CUDA and AMD ROCm. No cloud, no API keys, no data leaving your machine.
Why Run DS4 Locally?
- Full privacy — every token is generated on your hardware, so sensitive data never leaves your network.
- Zero recurring cost — after the initial setup there are no per-token fees, making heavy workloads dramatically cheaper.
- Cross-platform GPUs — the same engine runs on Macs (Metal), NVIDIA cards (CUDA) and AMD cards (ROCm).
- OpenAI-compatible API — drop it in behind existing tools that speak the OpenAI protocol.
Hardware Requirements
| Component | Minimum | Recommended |
|---|---|---|
| CPU | 2 cores | 4+ cores |
| RAM | 4 GB | 8+ GB |
| GPU | Metal / CUDA / ROCm capable | Apple Silicon, RTX or Radeon |
For a sense of real-world speed, antirez publishes benchmark charts from his own Apple Silicon testing:
Getting Started
The easiest path is the official Docker image arraying/ds4:latest (over 29,000 pulls). The container exposes a local API on port 8080 and stores model data under /data, so models are downloaded automatically on first start. The DS4 tool page includes a ready-to-use docker-compose.yml plus precise system requirements for planning your deployment.
🚀 Want to deploy DS4 yourself?
Docker configs, system requirements, and installation guides — all on one page.
View DS4 Tool Page →