# Qwen2.5-Coder 7B/14B/32B: Repo-Level Context (16K-128K) + FIM Native FT

> Source: https://sukruyusufkaya.com/learn/fine-tuning-cookbook/ftc-qwen-coder-repo-level-context
> Updated: 2026-08-13T12:22:15.667Z
> Category: Fine-Tuning Cookbook (Model-by-Model)
> Module: Part VIII — Code Models & Repo-Level FT
**TLDR:** Qwen2.5-Coder ailesi — 2025'in en güçlü açık code LLM. FIM native, 128K context, repo-level context için optimize. 32B HumanEval 92.7%, SWE-Bench-Lite 31.6%. RTX 4090'da 7B QLoRA 40 dk; 32B cloud H100 80GB tek-GPU.