# DeepSeek-R1-Distill (Llama-8B / Qwen-7B): Reasoning Trace Distillation — Learning 'Think Tokens'

> Source: https://sukruyusufkaya.com/en/learn/fine-tuning-cookbook/ftc-deepseek-r1-distill-reasoning
> Updated: 2026-05-14T14:42:51.952Z
> Category: Fine-Tuning Cookbook (Model-by-Model)
> Module: Part III — Small Open Models (1B–8B)
**TLDR:** DeepSeek-R1-Distill — Llama/Qwen bases distilled from R1 (671B) traces. <think>...</think> format, CoT trace dataset, compressing R1's reasoning into 7-8B. Your own reasoning FT on RTX 4090: 1000 R1 traces suffice.