Runs DeepSeek V4 Flash, GLM 5.x, and Qwen models locally on Metal, CUDA, or ROCm with a native engine, a CLI, a coding agent, and an OpenAI-compatible server.
Use it now
Keep it with up to 9 other tools you're setting up, then share them all as one link.
Try this first
Ask your coding agent:
- SSD Streaming
Use ds4 to stream model weights from SSD: ‹what you need›
Install the CLI
Install it as its own instructions say.
Its page: github.com/antirez/ds4
Call the API
Auth None needed
Its page: github.com/antirez/ds4
Compare
Press Compare to add it to the tray at the bottom of the window, then pick up to 2 others anywhere on the site.
- Access
- CLIAPI
- Auth
- None
- Skills
- 4
Skills
Local API Server
Serves OpenAI-style chat, Responses, completions, and Anthropic-style messages endpoints with tools and SSE streaming.
Native Coding Agent
Runs ds4-agent directly on the model without an HTTP server, using the model's native tool format and saved sessions.
SSD Streaming
Streams model weights from SSD so smaller Macs can run large models, such as full GLM 5.x on 128 GB systems.
Multi-GPU Serving
Uses multiple CUDA cards, including Ada Lovelace L40S, as a multi-user LLM server for the supported models.