Fine-Tuning a 3B Model to Internalize a 3500-Token SOP Cuts Costs 8.4x

Arpit Bhayani detailed his workflow-slm experiment, using QLoRA to compile a 3500-token incident-response SOP into a fine-tuned Qwen2.5 3B model, achieving 93% accuracy while cutting costs by 8.4x.

2026-10-24 ~ 2026-10-24 · 2 related posts

• Dev compiles incident-triage SOP into a fine-tuned 3B model: 93% accuracy at 8.4x lower cost — arpit_bhayani · 2026-10-24
• Write-up: compiling a 3,500-token workflow SOP into a QLoRA-tuned 3B model — arpit_bhayani · 2026-10-24