Can I run Qwen3 32B locally?
Yes. At Q4 with 8K context, Qwen3 32B needs about 21 GB, so it runs entirely on any GPU with 23 GB of VRAM or more, starting with the NVIDIA GeForce RTX 3090 Ti.
Memory needed by compression and context length
| Compression | 4K context | 8K context | 32K context |
|---|---|---|---|
Q4 The most common download. Small quality loss. | 20 GB | 21 GB | 27 GB |
Q5 Very close to full quality. | 24 GB | 25 GB | 31 GB |
Q8 Practically indistinguishable from FP16 at about half the size. | 34 GB | 35 GB | 41 GB |
FP16 Uncompressed. Reference quality, largest size. | 63 GB | 64 GB | 70 GB |
Total memory: model weights, context (KV cache) and about 1 GB of runtime overhead. To run entirely on the GPU, VRAM must be about 5% larger than this.
GPUs that run Qwen3 32B entirely on the GPU
- Amazon
NVIDIA GeForce RTX 3090 Ti
24 GB · ~34 tok/s at Q4
- Amazon
NVIDIA GeForce RTX 3090
24 GB · ~31 tok/s at Q4
- Amazon
NVIDIA GeForce RTX 4090
24 GB · ~34 tok/s at Q4
- Amazon
AMD Radeon RX 7900 XTX
24 GB · ~32 tok/s at Q4
- Amazon
NVIDIA GeForce RTX 4090 D
24 GB · ~34 tok/s at Q4
- Amazon
NVIDIA GeForce RTX 5090 D
32 GB · ~58 tok/s at Q4
- Amazon
NVIDIA GeForce RTX 5090
32 GB · ~58 tok/s at Q4
- Amazon
NVIDIA GeForce RTX 5090 D v2
32 GB · — at Q4
Model facts
- Parameters
- 32.8B
- Max context
- 32K tokens
- License
- Apache 2.0
- Released
- Apr 2025
Check it with your hardware
Change the compression, context length and system RAM in the calculator.
Open calculator