Open Model Training Prompt

A structured Claude Code prompt for fine-tuning and deploying local open-weight models

Description

A comprehensive prompt template that transforms Claude Code into an ML research lead and training engineer. It guides users through shortlisting models like Qwen 3.6-35B, setting up Colibri and RunPod, and executing SFT/LoRA/QLoRA experiments to achieve local model parity with Opus-level performance.

Descriptions, tags, and model credits may be AI-generated or inferred from public sources and can be incomplete or wrong. Learn more.

Details

Why this is interesting: Demonstrates a sophisticated one-shot workflow for complex ML engineering tasks using agentic coding tools.
Attributed to: Avid@Av1dlive
Built with: Codex Qwen3.6 Jev
Tags: #finetuning #claude-code #runpod #llm-ops

Discovered: X
First posted: X · Oct 3, 2026
Listed: Added just now
Build: Strong
Models: Moderate

Prompt

Act as my ML research lead and training engineer. Build an open-weight model setup I can run, improve, and own. Use Opus 3.5 as a comparison target. Measure capability; never promise parity.

Tasks: [what I want the model to do]
Hardware: [GPU / VRAM / RAM / storage]
Total budget: [amount] Deadline: [date]

Inspect what you can. Ask only for missing information that blocks progress.

Source