Oossa

DwarfStar releases ds4 for running large language models locally

The creator of Redis offers ds4, a C‑based engine that lets high‑memory Macs and GPU machines run DeepSeek, GLM and Qwen models on‑device.

NoteBy Published by Oossa: 1 min read

Antirez, the creator of Redis, announced ds4 on Oct 2 2026. The tool is a narrow C inference engine that runs frontier‑weight models like DeepSeek V4, GLM 5.x and Qwen 3.8 on high‑memory Macs, CUDA or ROCm machines. It uses asymmetric 2‑bit quantization and can stream KV cache to SSD, letting long prompts survive restarts. ds4 provides a CLI, a local HTTP API and a native coding agent that all share the same model state.

Why it matters

Developers can run powerful LLMs on their own hardware without relying on remote APIs, reducing latency and data‑privacy concerns.

Was this article useful?
Share

Read next

Oossa · Newsletter

The week in AI, explained

Every Monday: the stories worth knowing, in plain language. Free, no spam.