CanItRun Logocanitrun.

Qwen3 32B

Qwen3 32B needs roughly 23.9 GB VRAM at Q4_K_M quantization (75.0 GB at FP16). 74 GPUs we track can run it fully in VRAM at 8k context.

74 GPUs run this natively · 31 with CPU offload

Alibaba32.8B params128k contextApache 2.0Commercial use ok

Qwen3 32B is a 32.8B parameter dense model developed by Alibaba. Dense 32B model with thinking/non-thinking mode support.

To run Qwen3 32B locally: Q4_K_M's real total is ~23.9GB once KV cache and overhead are counted, just past a 24GB card's ~22.8GB effective capacity -- same tier as Qwen2.5-32B, both need CPU offload or a lower quant on a real 24GB card.

Chain-of-thought capabilities at 32B scale, strong reasoning with Apache 2.0 licensing.

VRAM at each quantization

Qwen3 32B natively supports a longer context window, but the table below is capped at 8k for comparability; KV cache grows linearly with context length.

QuantWeightsKV cacheTotal
FP32131.2 GB1.34 GB148.4 GB
BF1665.6 GB1.34 GB75.0 GB
FP1665.6 GB1.34 GB75.0 GB
Q8_034.9 GB1.34 GB40.5 GB
Q6_K26.9 GB1.34 GB31.7 GB
Q5_K_M23.4 GB1.34 GB27.7 GB
Q4_K_Mrec20.0 GB1.34 GB23.9 GB
Q3_K_M15.8 GB1.34 GB19.2 GB
Q2_K12.5 GB1.34 GB15.5 GB
NVFP4cuda16.4 GB1.34 GB19.9 GB

KV cache figures assume 8k context at FP16. NVFP4 quantization requires a CUDA-capable GPU. Enable TurboQuant in the calculator to see reduced KV cache estimates.

Benchmarks

MMLU-Pro
65.5

GPUs that run Qwen3 32B natively (74)

Show 69 more
Plus 31 GPUs that run it with CPU offload (slower)

Notes

Supports thinking (chain-of-thought) and non-thinking modes.

Hugging Face ↗Ollama ↗Released 2025-04-29

Compare Qwen3 32B with other models

Continue reading

vram-guides9 min
Best LLMs for 24 GB VRAM (2026)
vram-guides10 min
Qwen Family Guide: 2.5 to 3.6
hardware10 min
Best GPUs for Coding Agents in 2026
hardware9 min
Best GPUs $500-$1000 for LLMs (2026)
hardware10 min
NVIDIA RTX 50 Series (Blackwell) for LLMs: Complete Guide
tutorials10 min
LLM Benchmarks Explained: MMLU, HumanEval, SWE-bench, and More

Frequently asked questions

What are the VRAM requirements for Qwen3 32B?
Qwen3 32B requires approximately 23.9 GB of VRAM at Q4_K_M quantization, 40.5 GB at Q8, and 75.0 GB at FP16. These numbers assume 8k context window; VRAM scales linearly with context length due to the KV cache.
How many parameters does Qwen3 32B have?
Qwen3 32B has 32.8 billion parameters.
How capable is Qwen3 32B?
With an MMLU-Pro score of 65.54, Qwen3 32B delivers solid general-purpose performance suitable for most everyday tasks and professional use.
Can Qwen3 32B run on a 16 GB GPU?
No. At Q4_K_M, Qwen3 32B needs 23.9 GB of VRAM, more than 16 GB. You will need a 32 GB GPU like the RTX 5090.
Can Qwen3 32B run on a 24 GB GPU?
No. Even at Q4_K_M, Qwen3 32B needs 23.9 GB. Consider a 32 GB card like the RTX 5090.
What is the smallest quantization for Qwen3 32B that fits in 24 GB of VRAM?
At NVFP4, Qwen3 32B needs 19.9 GB, the highest-quality quantization that fits in 24 GB of VRAM.
What GPU do I need to run Qwen3 32B locally?
A 32 GB GPU is the minimum. At Q4_K_M, Qwen3 32B needs 23.9 GB VRAM. Good option: RTX 5090 (32 GB).