Serving local LLMs from an AMD GPU desktop to a Mac for vibe coding
DickIMeanRichard · reddit · 2026-07-26
A veteran IT professional shared their cross-device local LLM deployment setup aimed at vibe coding and automation.
- Hardware & Network: Uses a MacBook Pro (16GB RAM) as the frontend, connecting via LAN to a Windows desktop with an AMD 7800XT GPU (16GB VRAM) and 64GB RAM. The desktop operates headlessly, dedicated to AI inference.
- Model Selection: The laptop runs smaller Qwen and Gemma models for light tasks; the desktop runs Qwen 35B and 27B models via Llama.cpp for heavy coding and debugging.
- Goals & Discussion: The author uses this setup for personal app dev, N8N automation, and NAS management, seeking community advice on model compatibility, MCP integration, and architecture optimization.
More from coding & agent
- Comet adds an automated debugging agent for Opik that scans ClickHouse traces — dl_weekly · 2026-07-26
- An agentic AI engineer maps out a full modern MLOps and coding stack — kmeanskaran · 2026-07-26
- Anthropic users say Opus 5 is best for complex agents, while Sonnet 5 fits routine coding — dr_cintas · 2026-07-26
- Claude Code sessions feel like different siblings, one user says — code_star · 2026-07-26
- SEOS 3.0.0 launches as a local-first multi-agent system for software engineering — Consistent-Gold-425 · 2026-07-26
- Grok Build should get a Cursor-style GUI before it can compete with Codex — mark_k · 2026-07-26