Single RTX 3090 + Qwen 3 8B with deepseek harness 'game changing', dev reports
tlpta · reddit · 2026-09-18
In a follow-up to a recent求助-style thread, the author reports a breakthrough local coding setup: switching to Qwen 3 8B (Q4, 256k context, 32k default gen length — kobold's cap) and moving from opencode to the deepseek harness. On a single RTX 3090, the author calls the result shockingly game-changing and finally understands the local coding-agent hype — a directly reproducible setup for similar hardware.
More from coding & agent
- Dev bets OpenAI DevDay unveils a unified persistent agent in the mold of Muse — brandon_galang · 2026-09-18
- Perplexity rolls out effort controls in Computer; user shares a 3-tier model workflow — inductionheads · 2026-09-18
- TypeSafe's structured evaluation model Jev goes live on Cloudflare AI Gateway — michellechen · 2026-09-18
- SVG Animations to Explain How an Event Loop Drives a Coding Agent — blaizedsouza · 2026-09-18
- Arize's Code Mode: Let Agents Write Code to Transform Data Outside the LLM Context — aparnadhinak · 2026-09-18
- One Prompt Bypassed Claude Code's Guardrails, Sparking a Call for Visible Safety Controls — avt_im · 2026-09-18