LLM GPU requirements
Qwen3 32B GPU requirements
How much GPU memory Qwen3 32B needs in BF16, FP8 and 4-bit, how its KV cache grows with context, what fine-tuning takes, and which servers run it, with monthly prices.
- Parameters
- 32.8B
- Architecture
- Dense 64 layers
- KV heads
- 8 × 128 dim
- Context
- 32k tokens
Cheapest servers that run it
FP8 weights · 8k context · 8 concurrent requests · about 54 GB
-
NVIDIA RTX PRO 6000 Blackwell In stock
96 GB GDDR7 · Single GPU
$1,390/mo
-
NVIDIA H100 SXM In stock
80 GB HBM3 · Single GPU
$2,190/mo
-
NVIDIA H200 In stock
141 GB HBM3e · Single GPU
$3,090/mo
Delivery date confirmed before you pay. Missed it? Full refund.
Inference memory
How much VRAM Qwen3 32B needs
Weights take parameters × bytes per parameter. Serving adds the KV cache and about 10% runtime overhead; the totals below assume an 8k context with 8 concurrent requests and a BF16 KV cache.
| Precision | Weights | Total to serve | Cheapest server that fits |
|---|---|---|---|
| BF16 2 bytes per parameter | 66 GB | 90 GB | NVIDIA H200 $3,090/mo |
| FP8 1 byte per parameter | 33 GB | 54 GB | NVIDIA RTX PRO 6000 Blackwell $1,390/mo |
| 4-bit 0.5 bytes per parameter | 16 GB | 36 GB | NVIDIA L40S $890/mo |
Estimates with about 8% of GPU memory kept free. Quantized weights also keep some layers in higher precision, and frameworks differ in overhead; check the free memory your engine reports at startup. A model split across GPUs with tensor parallelism runs best on NVLink servers (HGX); the 8× RTX PRO 6000 server links its GPUs over PCIe.
KV cache
Qwen3 32B KV cache by context length
Each token in flight stores keys and values for every layer: 2 × 64 layers × 8 KV heads × 128 dimensions × 2 bytes = 256 KB per token in BF16. An FP8 KV cache halves these figures.
| Context per request | 1 request | 8 requests | 32 requests |
|---|---|---|---|
| 4k tokens | 1.1 GB | 8.6 GB | 34 GB |
| 32k tokens | 8.6 GB | 69 GB | 275 GB |
| 128k tokens | 34 GB | 275 GB | 1,100 GB |
Fine-tuning
Memory to fine-tune Qwen3 32B
Rough totals before tuning batch size and sequence length: QLoRA keeps 4-bit base weights, LoRA keeps BF16 weights frozen, and a full fine-tune with AdamW needs about 16 bytes per parameter plus activations, sharded across GPUs.
| Method | Approximate memory | Cheapest server that fits |
|---|---|---|
| QLoRA (4-bit base) | 25 GB | NVIDIA GeForce RTX 5090 $449/mo |
| LoRA (BF16 base) | 81 GB | NVIDIA RTX PRO 6000 Blackwell $1,390/mo |
| Full fine-tune (AdamW, mixed precision) | 630 GB | 8× NVIDIA RTX PRO 6000 Blackwell $10,490/mo |
Full fine-tuning assumes sharded training (FSDP or ZeRO-3) with activation checkpointing. See the fine-tuning guide for methods and settings.
About Qwen3 32B
Qwen3 32B is the largest dense model of the Qwen3 family: 64 layers, 8 KV heads and Apache 2.0 weights. With 64 layers against 40 for Qwen3 14B, its KV cache per token is 60% larger, so long contexts and high concurrency weigh heavily in your sizing.
Every figure on this page comes from the model’s published architecture and the same formulas as our VRAM calculator, where you can change the context length, concurrency and KV cache precision.
Model facts
- Developer
- Alibaba Qwen
- License
- Apache 2.0
- Type
- Dense
- Parameters
- 32.8B
- Layers
- 64
- Attention
- 64 query heads, 8 KV headsgrouped-query attention, head dimension 128
- Context window
- 32k tokens native, about 128k with YaRN
How much VRAM does Qwen3 32B need?
About 66 GB for the weights in BF16, 33 GB in FP8 and 16 GB in 4-bit, plus the KV cache: about 0.3 GB per 1,000 tokens in flight with a BF16 cache. Serving it in FP8 with an 8k context and 8 concurrent requests takes about 54 GB in total.
Can Qwen3 32B run on a single GPU?
Yes. In FP8, a single NVIDIA RTX PRO 6000 Blackwell (96 GB) serves it with an 8k context and 8 concurrent requests. Longer contexts or more users need more memory for the KV cache, so check your case in the VRAM calculator.
What is the cheapest FullGPU server for Qwen3 32B?
For FP8 serving with an 8k context and 8 concurrent requests, the NVIDIA RTX PRO 6000 Blackwell at $1,390 per month is the lowest-priced configuration that fits. For other contexts and concurrency, size it in the VRAM calculator or ask an engineer.
How much memory do I need to fine-tune Qwen3 32B?
Roughly 25 GB with QLoRA, 81 GB with LoRA on BF16 weights, and 630 GB for a full fine-tune with AdamW in mixed precision, before tuning batch size and sequence length. Full fine-tunes of large models are sharded across GPUs with FSDP or ZeRO-3.
Talk to an engineer
Run Qwen3 32B on your own GPUs.
Tell us what you need. A GPU engineer replies with availability, a confirmed delivery date and a payment link.
- Delivery date confirmed before you pay
- Cancel any month
- Card or crypto, same price