QwQ 32B VRAM requirements
QwQ 32B has 32.76 billion parameters in its published weight files. At 4 bits, the weights alone need approximately 16.4 GB. The default inference scenario estimates 20.9 GB including cache and runtime headroom.
Checked 20 October 2026. Memory estimates use decimal GB. Quantized weights may add scales, metadata and buffers; a compatible implementation and model licence are still required.
Change the memory assumptions
Capacity line: 24 GB, from the lowest-priced single-GPU candidate. Memory fit alone does not establish compatibility.
Memory at different precisions
| Precision | Weights GB | Inference GB | LoRA scenario GB | Full fine-tune GB |
|---|---|---|---|---|
| 16-bit | 65.5 | 79.9 | 73.4 | 591.7 |
| 8-bit | 32.8 | 40.6 | 40.7 | 591.7 |
| 4-bit | 16.4 | 20.9 | 24.3 | 591.7 |
How the estimate is calculated
Weights = parameters × bits ÷ 8. For supported standard attention, KV cache = 2 × layers × KV heads × head dimension × context tokens × batch size × 2 bytes. Inference adds the chosen percentage of runtime headroom to weights and cache. Divide bytes by 1,000,000,000 for GB.
The LoRA scenario keeps the chosen base weights and allows 18 bytes per trainable adapter parameter, plus the activation allowance. Full fine-tuning uses 18 bytes per parameter for a mixed-precision AdamW scenario, plus that allowance. Its total does not fall when you select 4-bit inference weights. These are budget estimates, not measured memory peaks; training context, optimizer, checkpointing and implementation can change them substantially.
Single-GPU candidates for this inference scenario
The offers update with the memory inputs above. Availability not reported means stock has not been confirmed. Memory capacity alone does not establish runtime compatibility or speed.
More filters: provider, memory, node size, stock, place, sort, price unit
“Machines in” filters the 4 of 60 offers whose provider says where the machine is; the others are under “Place not stated”.
CoreWeave, DigitalOcean, Lambda, Massed Compute, OVHcloud and Thunder Compute put GPUs in North America on their own pages, but none of their offers here says so.
| GPU / provider | Billing | $/GPU-hour | Node · total per hour | Rent |
|---|---|---|---|---|
| Tesla V100 · Verda (DataCrunch)16 GB | on-demand | $0.23+26.7% in 30 days | 1× $0.23 | Rent → |
| A30 · Massed Compute24 GB | on-demand | $0.35 | 1× $0.35 | Rent → |
| RTX A6000 · Thunder Compute48 GB | on-demand | $0.35 | 1× $0.35 | Rent → |
| RTX A5000 · Massed Compute24 GB | on-demand | $0.44 | 1× $0.44 | Rent → |
| RTX A6000 · Massed Compute48 GB | on-demand | $0.57 | 1× $0.57 | Rent → |
| Quadro RTX 5000 · OVHcloud16 GB | on-demand | $0.60 | 1× $0.60 | Rent → |
| Quadro RTX 6000 · Lambda24 GB · 1x NVIDIA Quadro RTX 6000 | on-demand | $0.69 | 1× $0.69 | Rent →checked 22 OctDetails
|
| RTX 4000 Ada · DigitalOcean20 GB · gpu-rtx4000x1-20gb | on-demand | $0.76 | 1× $0.76 | Rent → |
| Tesla V100 · OVHcloud16 GB | on-demand | $0.77 | 1× $0.77 | Rent → |
| RTX A6000 · Verda (DataCrunch)48 GB | on-demand | $0.78+27.9% in 30 days | 1× $0.78 | Rent → |
| RTX 6000 Ada · Massed Compute48 GB | on-demand | $0.79 | 1× $0.79 | Rent → |
| L40 · Thunder Compute48 GB | on-demand | $0.79 | 1× $0.79 | Rent → |
| L4 · AWS24 GB · g6.xlarge | on-demand | $0.80 | 1× $0.80 | Rent →checked 22 OctDetails
|
| L40 · Massed Compute48 GB | on-demand | $0.86 | 1× $0.86 | Rent → |
| RTX PRO 4500 Blackwell · Massed Compute32 GB | on-demand | $0.92 | 1× $0.92 | Rent → |
| L40S · Massed Compute48 GB | on-demand | $0.97 | 1× $0.97 | Rent → |
| L4 · OVHcloud24 GB | on-demand | $1.00 | 1× $1.00 | Rent → |
| A10 · OVHcloud24 GB | on-demand | $1.00 | 1× $1.00 | Rent → |
| RTX A6000 · Lambda48 GB · 1x NVIDIA A6000 | on-demand | $1.09 | 1× $1.09 | Rent →checked 22 OctDetails
|
| A100 · Thunder Compute80 GB | on-demand | $1.09 | 1× $1.09 | Rent → |
| A10 · Lambda24 GB · 1x NVIDIA A10 | on-demand | $1.29 | 1× $1.29 | Rent →checked 22 OctDetails
|
| A100 · Massed Compute80 GB | on-demand | $1.35 | 1× $1.35 | Rent → |
| A100 SXM 80GB · Massed Compute80 GB | on-demand | $1.38 | 1× $1.38 | Rent → |
| RTX 6000 Ada · Verda (DataCrunch)48 GB | on-demand | $1.40+25.9% in 30 days | 1× $1.40 | Rent → |
| A100 SXM 40GB · Verda (DataCrunch)40 GB | on-demand | $1.55+21.7% in 30 days | 1× $1.55 | Rent → |
| RTX 6000 Ada · DigitalOcean48 GB · gpu-rtx6000x1-48gb | on-demand | $1.57 | 1× $1.57 | Rent → |
| L40S · DigitalOcean48 GB · NVIDIA L40S 1x | on-demand | $1.57 | 1× $1.57 | Rent → |
| A40 · Vultr48 GB | on-demand | $1.71 | 1× $1.71 | In stockRent → |
| L40S · OVHcloud48 GB | on-demand | $1.80 | 1× $1.80 | Rent → |
| L40S · Verda (DataCrunch)48 GB | on-demand | $1.81+23.2% in 30 days | 1× $1.81 | Rent → |
| L40S · AWS48 GB · g6e.xlarge | on-demand | $1.86 | 1× $1.86 | Rent →checked 22 OctDetails
|
| A100 SXM 40GB · Lambda40 GB · 1x NVIDIA A100 SXM | on-demand | $1.99 | 1× $1.99 | Rent →checked 22 OctDetails
|
| A100 PCIe 40GB · Lambda40 GB · 1x NVIDIA A100 PCIe | on-demand | $1.99 | 1× $1.99 | Rent →checked 22 OctDetails
|
| A100 SXM 80GB · Verda (DataCrunch)80 GB | on-demand | $2.11+21.4% in 30 days | 1× $2.11 | Rent → |
| RTX PRO 6000 · Massed Compute96 GB | on-demand | $2.19 | 1× $2.19 | Rent → |
| GH200 · Lambda96 GB · 1x NVIDIA GH200 | on-demand | $2.29 | 1× $2.29 | Rent →checked 22 OctDetails
|
| RTX PRO 6000 · Verda (DataCrunch)96 GB | on-demand | $2.47+32.0% in 30 days | 1× $2.47 | Rent → |
| MI300X · DigitalOcean192 GB · gpu-mi300x1-192gb | on-demand | $2.59 | 1× $2.59 | Rent →checked 22 OctDetails
|
| H100 · Massed Compute80 GB | on-demand | $2.73 | 1× $2.73 | Rent → |
| H100 · OVHcloud80 GB | on-demand | $2.99 | 1× $2.99 | Rent → |
| A100 · OVHcloud80 GB | on-demand | $3.07 | 1× $3.07 | Rent → |
| H100 SXM · Massed Compute80 GB | on-demand | $3.14+8.7% in 30 days | 1× $3.14 | Rent → |
| H100 SXM · Thunder Compute80 GB | on-demand | $3.20 | 1× $3.20 | Rent → |
| H100 PCIe · Lambda80 GB · 1x NVIDIA H100 PCIe | on-demand | $3.29 | 1× $3.29 | Rent →checked 22 OctDetails
|
| H200 NVL · Massed Compute141 GB | on-demand | $3.62 | 1× $3.62 | Rent → |
| A100 PCIe 80GB · Microsoft Azure80 GB · Standard_NC24ads_A100_v4 | on-demand | $3.67 | 1× $3.67 | Rent →checked 22 OctDetails
|
| MI325X · DigitalOcean256 GB · gpu-mi325x1-256gb | on-demand | $3.80 | 1× $3.80 | Rent →checked 22 OctDetails
|
| H100 SXM · Lambda80 GB · 1x NVIDIA H100 SXM | on-demand | $4.29 | 1× $4.29 | Rent →checked 22 OctDetails
|
| H100 · DigitalOcean80 GB · gpu-h100x1-80gb | on-demand | $4.41 | 1× $4.41 | Rent →checked 22 OctDetails
|
| H100 SXM · Verda (DataCrunch)80 GB | on-demand | $4.47+32.8% in 30 days | 1× $4.47 | Rent → |
| H200 (SXM) · DigitalOcean141 GB · gpu-h200x1-141gb | on-demand | $4.47 | 1× $4.47 | Rent →checked 22 OctDetails
|
| H200 (SXM) · Verda (DataCrunch)141 GB | on-demand | $5.83+32.1% in 30 days | 1× $5.83 | Rent → |
| MI350X · DigitalOcean288 GB · gpu-mi350x1-288gb | on-demand | $6.16 | 1× $6.16 | Rent →checked 22 OctDetails
|
| GH200 · CoreWeave96 GB | on-demand | $6.50 | 1× $6.50 | Rent → |
| H100 NVL · Microsoft Azure94 GB · Standard_NC40ads_H100_v5 | on-demand | $6.98 | 1× $6.98 | Rent →checked 22 OctDetails
|
| B200 · Lambda180 GB · 1x NVIDIA B200 SXM6 | on-demand | $6.99 | 1× $6.99 | Rent →checked 22 OctDetails
|
| B200 · Verda (DataCrunch)180 GB | on-demand | $8.27+26.3% in 30 days | 1× $8.27 | Rent → |
| B300 · Verda (DataCrunch)268 GB | on-demand | $10.58+32.9% in 30 days | 1× $10.58 | Rent → |
| B300 · DigitalOcean288 GB · gpu-b300x1-288gb | on-demand | $11.19 | 1× $11.19 | Rent →checked 22 OctDetails
|
| GB300 · Verda (DataCrunch)288 GB | on-demand | $12.17+32.9% in 30 days | 1× $12.17 | Rent → |
Try another provider or reduce the minimum memory.
Tables rank by price, lowest first. Commercial relationships never affect ordering. How we collect and rank prices.
Same GPU, provider, billing and price on several hosts or listings: one row, with the count. – means the provider does not state it.
Monthly view assumes 730 hours of continuous use. Daily view assumes 24 hours.
Cite this data
GPU Rental Prices, "QwQ 32B VRAM: memory estimates and GPU rental costs", prices checked 22 October 2026, https://gpurentalprices.com/models/qwq-32b
Quote it with a link to this page. Downloads of the whole dataset are on Data & API.
Price a training run of QwQ 32B and check it fits the node · Compare an API bill with self-hosting
Other models from this publisher
- Qwen2.5 0.5B
- Qwen2.5 1.5B
- Qwen2.5 3B
- Qwen2.5 7B
- Qwen2.5 14B
- Qwen2.5 32B
- Qwen2.5 72B
- Qwen2.5 Coder 0.5B
- Qwen2.5 Coder 1.5B
- Qwen2.5 Coder 3B
- Qwen2.5 Coder 7B
- Qwen2.5 Coder 14B
- Qwen2.5 Coder 32B
- Qwen3 0.6B
- Qwen3 1.7B
- Qwen3 4B
- Qwen3 8B
- Qwen3 14B
- Qwen3 32B
- Qwen3 30B A3B
- Qwen3 235B A22B
- Qwen3 4B 2507
- Qwen3 30B A3B 2507
- Qwen3 235B A22B 2507
- Qwen3 Coder 30B A3B
- Qwen3 Coder 480B A35B
- Qwen3 Next 80B A3B