Model

QwQ-32B

OPEN

Alibaba · QwQ family · released Mar 6, 2025

RL-scaled reasoning flagship matching DeepSeek-R1 (671B) at 1/20th the size; Apache 2.0.

ReasoningCodingVisionFunction callingTool useAgentic

1260.2

Elo · rank #165

Parameters

32.5B

Active params

32.5B (dense)

Context

128K tokens

Architecture

Dense Transformer (GQA, RL-scaled reasoning)

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

AIMEMath79.5%#81

best: GPT-5.2 · 100.0%

Arena-HardHuman preference89.5%#12

best: Qwen3-235B-A22B (Non-Thinking) · 96.1%

BIG-Bench HardReasoning29.5%#135

best: ERNIE 4.5 300B-A47B · 94.3%

BFCL v3Agents66.4%#19

best: Hunyuan-A13B · 78.3%

C-EvalKnowledge88.4%#15

best: Qwen3.6-Plus · 93.3%

CodeforcesCoding1982#21

best: DeepSeek-V4-Pro (Think Max) · 3206

GPQA DiamondReasoning65.6%#143

best: GPT-5.6 · 94.6%

IFEvalReasoning83.9%#72

best: Gemma 4 26B A4B · 98.5%

LiveCodeBenchCoding63.4%#92

best: DeepSeek-V4-Pro (Think Max) · 93.5%

MATH-500Math98.0%#9

best: GPT-5 · 99.4%

MMLU-ProKnowledge69.1%#93

best: Claude Fable 5 · 91.5%

MMLU-ReduxKnowledge90.0%#22

best: Qwen3.7-Max · 95.0%

Run it locally

VRAM @ Q4

19.85 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

Throughput data unavailable.

Quantizations

GGUF · AWQ

QwQ family

Elo progression across releases

QwQ-32B-PreviewNov 20241080.0 QwQ-32BMar 20251260.2

Resources

HFQwen/QwQ-32B GHQwenLM/QwQ DOCSqwenlm.github.io/blog/qwq-32b APIAvailable via Alibaba API

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