Qwen-Image · 4 GPUs measured first-party · text-to-image · Updated October 2026
What GPU Do You Need for Qwen-Image?
Qwen-Image on 4 GPUs, measured first-party: NVIDIA H200 leads at 3.14 images/min, A100 80GB SXM4 trails at 1.34 images/min, and it peaked at 63GB of VRAM.
Benchmarked weights: Qwen/Qwen-Image
NVIDIA H200
3.14 images/min on Qwen-Image, the ceiling. Measured on our bench. 141GB of VRAM, $31,000 at launch.
- 3.14 images/min on Qwen-Image
- 141GB, clears the Qwen-Image floor
- Rentable by the hour rather than bought
- 700W board rating
- Datacenter or workstation hardware, not a retail purchase
NVIDIA RTX PRO 6000 Blackwell Workstation Edition
1.97 images/min on Qwen-Image, lowest launch price that still fits. Measured on our bench. 96GB of VRAM, $8,565 at launch.
- 1.97 images/min on Qwen-Image
- 96GB, clears the Qwen-Image floor
- Rentable by the hour rather than bought
- 600W board rating
- Datacenter or workstation hardware, not a retail purchase
What GPU Do You Need for Qwen-Image?, images/min by GPU
Efficiency: images/min per 100W drawn
Power is the average pulled during the run, sampled at 1Hz. The fastest card is often not the one here, and for anything left running this is the number that shows up on the bill.
Value: images/min per $1,000 of MSRP
Launch price, not street price, so it ages. A speed leaderboard always crowns the most expensive card; this is the counterweight.
Qwen-Image. Measured image generation speed by GPU
| GPU | Images/min | s per image | img/W·min | Avg power |
|---|---|---|---|---|
| NVIDIA H200 | 3.14 | 19.14 | 0.005 | 686.1 W |
| NVIDIA H100 80GB HBM3 | 3 | 19.99 | 0.004 | 679.0 W |
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 1.97 | 30.53 | 0.004 | 519.0 W |
| NVIDIA A100 80GB SXM4 | 1.34 | 44.84 | 0.003 | 395.8 W |
What the numbers show. Across 4 GPUs measured on our own bench, H200 is fastest at 3.14 images/min. The slowest, A100 80GB SXM4, manages 1.34, so the spread is 2.3x from top to bottom. Per dollar of launch price, RTX PRO 6000 Blackwell Workstation Edition gives the most (0.2 images/min per $1,000).
About Qwen-Image. Qwen-Image: from Qwen, 20B parameters, on Hugging Face since August 2025, Apache 2.0 licence. 404,854 downloads in the last 30 days and 5 community quantizations.
How it compares. H100 80GB HBM3: Qwen-Image 3.0 images/min, Qwen-Image-Edit 3.68 (20B), Qwen-Image 2512 3.0 (20B), FLUX.1 dev 9.06 (12B), FLUX.1 Schnell 58.36 (12B). 3 of 4 beat Qwen-Image here.
Cost on a rented GPU. 1,000 images of Qwen-Image: $9.10 on a RTX PRO 6000 Blackwell Workstation Edition ($1.08/hr, 8.5 hours), $19.06 on a H200 ($3.59/hr, 5.3 hours, 2.1x the cost).
Qwen-Image: cost per 1,000 images on rented GPUs
| GPU | Cheapest rate | Speed (images/min) | Cost per 1,000 images |
|---|---|---|---|
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | $1.08/hr | 1.97 | $9.10 |
| NVIDIA A100 80GB SXM4 | $0.95/hr | 1.34 | $11.78 |
| NVIDIA H100 80GB HBM3 | $2.14/hr | 3 | $11.87 |
| NVIDIA H200 | $3.59/hr | 3.14 | $19.06 |
Cheapest hourly rate we track on RunPod and Vast.ai, divided by the measured speed. Startup time and storage are extra.
Speed tiers for Qwen-Image. under 6 images/min: 4 (H200, H100 80GB HBM3, RTX PRO 6000 Blackwell Workstation Edition). 30 images/min means two seconds or less per picture.
Time per image. Qwen-Image at 1024px and 50 steps: 19.1s per image on the H200, 44.8s on the A100 80GB SXM4. A batch of 100 takes 32 min on the fastest card and 75 min on the slowest.
VRAM for Qwen-Image. Measured peak 59.3GB, so 80GB is the smallest common card size; smallest card it ran on: H100 80GB HBM3 (80GB).
Power on Qwen-Image. Most efficient: H200, 686W, 3.64 kWh per 1,000 images. At $0.15/kWh: $0.55 per 1,000 images.
Our verdict
Fastest on Qwen-Image: NVIDIA H200, 3.14 images/min. Cheapest to rent per job: NVIDIA RTX PRO 6000 Blackwell Workstation Edition, $9.10 per 1,000 images.
FAQ
- What GPU do I need to run Qwen-Image?
- About 62GB. Smallest card that ran it: NVIDIA H100 80GB HBM3 (80GB).
- How much does it cost to run Qwen-Image in the cloud?
- $9.10 per 1,000 images on a NVIDIA RTX PRO 6000 Blackwell Workstation Edition at $1.08/hr, cheapest of 4 rentable cards we measured.
- Can I run Qwen-Image on a 12GB, 16GB or 24GB card?
- It used 59.3GB at the precision we tested. 12GB: no; 16GB: no; 24GB: no.
- Is the H100 80GB HBM3 or the A100 80GB SXM4 faster for Qwen-Image?
- The H100 80GB HBM3: 3.0 vs 1.34 images/min, 124% faster on our bench.
How we test
Qwen-Image at 1024px in diffusers, bf16, native precision with no offload, timed over three generations after a warmup, with power and VRAM sampled throughout. Image speed tracks tensor compute and architecture generation more than memory bandwidth, so the order here differs from our LLM boards.