Running a minimal local agent on 24GB RAM: Pi Agent + Bonsai 2 9B hits ~30 tok/s
Potential_Block4598 · reddit · 2026-11-12
A Redditor built a deliberately minimal local agentic stack on a secondary workstation (24GB RAM, 12GB VRAM): Pi Agent + Bonsai 2 9B + a SearXNG search extension.
- Reports 30 tok/s and performance 'unbelievably better than expected'
- Prioritizes simplicity and speed over capability: no heavy frameworks, no long prompts, the simplest search extension available
- Models above 9B run too slow; tried nanbeige but found it overthinks
- Considering but deferring context compression, sub-agent delegation, sandboxed memory and a Ralph-loop to-do list; soliciting model suggestions
More from coding & agent
- Game dev workflow: AI-generated sprite sheets turned into character walk GIFs — algo_diver · 2026-11-12
- Agent reads email threads and pre-books your apartment viewing on Gcal automatically — braelyn_ai · 2026-11-12
- AntiFox AI Pentesting Platform Linked to Attacks That Stole AWS Production Secrets — Thionne_WTZ · 2026-11-12
- Dev builds npx driftcost CLI to check if Claude Code token usage silently shifted between Opus versions — Tough_Stay8909 · 2026-11-12
- AFK Pilot ships standalone desktop app to run coding agents without an IDE — PawelHuryn · 2026-11-12
- Gemini 4 Carbon now usable in VS Code and Cursor via AFK Pilot extension — PawelHuryn · 2026-11-12