Model

Qwen2.5-Coder-32B

OPEN

Alibaba · Qwen2.5-Coder family · released Nov 12, 2024

Apache-2.0 coding flagship; Instruct sibling rivals GPT-4o on EvalPlus/LiveCodeBench/BigCodeBench and Aider code-repair.

ReasoningCodingVisionFunction callingTool useAgentic

803.8

Elo · unrated

Parameters

32.5B

Active params

32.5B (dense)

Context

128K tokens

Architecture

Dense transformer, 64 layers (GQA 40Q/8KV)

License

Apache-2.0

Languages

—

API price (in/out)

No hosted API

Modalities

text

Benchmark results

Bar shows position within the tracked field; marker = field best

BigCodeBenchCoding53.6%#2

best: GPT-4o mini · 57.4%

HumanEval FIMCoding88.3%#2

best: Codestral 22B · 91.6%

HumanEval+Coding60.4%#24

best: Mistral Small 3.2 (24B) · 92.9%

HumanEvalCoding65.9%#90

best: Claude Opus 4.5 · 99.4%

MBPP+Coding68.2%#18

best: Llama 3.1 405B · 88.6%

MBPPCoding83.0%#10

best: Llama-3.3-Nemotron-Super-49B v1 (Reasoning On) · 91.3%

Run it locally

VRAM @ Q4

19.9 GB

VRAM @ FP16

65 GB

Fits on (Q4)

RTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB

~20GB Q4 fits within 24GB with limited headroom for context/KV cache; no single independently verified 4090 tok/s figure found.

Quantizations

GGUF · AWQ · GPTQ Int4 · MLX

Qwen2.5-Coder family

Elo progression across releases

Qwen2.5-Coder-32BNov 2024803.8

Resources

HFQwen/Qwen2.5-Coder-32B GHQwenLM/Qwen2.5-Coder DOCSqwenlm.github.io/blog/qwen2.5-coder-family

API price weights · each benchmark row carries its own source badge (see methodology)