Qwen2.5-Coder-32B locally: a private Copilot on a 24GB card
Qwen2.5-Coder-32B still runs a serious private coding assistant on a single 24GB card, with a 32K native context (128K with YaRN). The VRAM math,…
Aliteq
Lena Fischer · AI & Local Compute Editor
The short answer
Qwen2.5-Coder-32B is a strong open coding model that runs a private, offline Copilot on a single 24GB card. At Q4_K_M our VRAM engine computes about 23 GB — a tight but real fit for an RTX 3090 or…
32.5B params, dense, 32K native context (128K with YaRN), Apache-2.0 — from the model card (Nov 2024).
Q4_K_M ~23.3 GB (computed by our VRAM engine, 16K context) — a tight fit on a 24GB card.
Run qwen2.5-coder:32b explicitly; the newer Qwen3-Coder is the MoE option.
Aliteq
Read the full story
Qwen2.5-Coder-32B locally: a private Copilot on a 24GB card