LOCAL_AI_STACK
v2.2 · 2026-06-12
← back to models

Qwen2.5-Coder 32B Instruct

qwen2-5-coder-32b-instruct
CODE

Qwen2.5-Coder 32B Instruct is a 32.5B code model profile for local AI planning. This page records provider, license, context, quantization, hardware-fit estimates, setup hints, source links, and caveats so readers do not choose a model by name alone.

PROVIDER
Qwen
FAMILY
Qwen2.5-Coder
MODEL TYPE
code
PARAMETERS
32.5B
MODALITIES
text
ARCHITECTURE
Decoder-only causal Transformer using Qwen2 architecture with RoPE, SwiGLU, RMSNorm, Attention QKV bias, and grouped-query attention; code-specific Qwen2.5-Coder post-training.
CONTEXT WINDOW
131,072 tokens
TRAINING TOKENS
5.5T
RELEASE DATE
2024-11-12
LICENSE & LINKS
GGUF REPOSITORIES
Official Qwen GGUF quantization repository; large variants may be split into multiple GGUF shards for a single quantization.
SETUP HINTS
OLLAMA
Use `ollama run qwen2.5-coder:32b` for Ollama's packaged build, or `ollama run hf.co/Qwen/Qwen2.5-Coder-32B-Instruct-GGUF:Q4_K_M` to pin the official Hugging Face GGUF.
LM STUDIO
Search for `Qwen/Qwen2.5-Coder-32B-Instruct-GGUF` in LM Studio and select Q4_K_M or Q5_K_M.
LLAMA.CPP
Use `llama-server -hf Qwen/Qwen2.5-Coder-32B-Instruct-GGUF:Q4_K_M`; for split GGUF downloads, download all shards for the chosen quantization and merge if your runtime requires a single file.
QUANTIZATION FILE SIZES (GGUF)
Q4_K_M
19.9 GB
Q5_K_M
23.3 GB
Q8_0
34.8 GB
Quantization note: Q4_K_M for broad local compatibility; Q5_K_M for better code quality if memory allows.
RAM / VRAM ESTIMATES
MIN RAM
24.9 GB
COMFORTABLE RAM
31.5 GB
MIN VRAM
23.9 GB
COMFORTABLE VRAM
30.3 GB
These are conservative local-inference estimates, not official hardware requirements.
HARDWARE FIT
CPU only (no GPU)
possible_but_slow
8 GB RAM
not_recommended
16 GB RAM
not_recommended
32 GB RAM
good
8 GB VRAM
not_recommended
12 GB VRAM
not_recommended
24 GB VRAM
usable
Apple Silicon (unified memory)
good_32gb_plus
Hardware fit values are conservative local-inference estimates based on GGUF size plus runtime and KV-cache overhead. Actual requirements depend on context length, quantization, runtime, GPU offload, and other running apps.
BEST FOR
·High-quality local coding, code repair, code reasoning, and programming-assistant workflows.
·Aider-style repair tasks and multi-file reasoning when hardware is sufficient.
·Developers who want the strongest local Qwen2.5 coding record in this batch.
AVOID IF
·You have less than 32GB system RAM or limited disk bandwidth.
·You need a lightweight always-on background coding assistant.
·You want a general chat model more than a code-specialist model.
CAVEATS
·RAM and VRAM figures are estimates from GGUF file size plus conservative runtime and KV-cache overhead; exact memory depends on context length, batch size, backend, and GPU offload.
·Official Qwen2.5-Coder cards state long-context support up to 128K, but GGUF cards and local runtimes commonly default to 32K unless configured otherwise.
·Tool-use support was not treated as verified for this record because the official model card describes code-agent use cases but does not provide a specific tool-calling support field for this checkpoint.
SOURCE URLS
FIELD EVIDENCE
LANGUAGES
40+ programming languages
CAPABILITIES
code tuned
← Qwen2.5-Coder 7B InstructQwen3 4B →