Qwen-Image-Edit 2509 · 4 GPUs measured first-party · text-to-image · Updated October 2026
Qwen-Image-Edit 2509 on 4 GPUs, measured first-party: NVIDIA H200 leads at 1.71 images/min, A100 80GB SXM4 trails at 0.75 images/min, and it peaked at 63GB of VRAM.
Benchmarked weights: Qwen/Qwen-Image-Edit-2509

1.71 images/min on Qwen-Image-Edit 2509, the ceiling. Measured on our bench. 141GB of VRAM, $31,000 at launch.

1.13 images/min on Qwen-Image-Edit 2509, lowest launch price that still fits. Measured on our bench. 96GB of VRAM, $8,565 at launch.
What GPU Do You Need for Qwen-Image-Edit 2509?, 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-Edit 2509. Measured image generation speed by GPU
| GPU | Images/min | s per image | img/W·min | Avg power |
|---|---|---|---|---|
| NVIDIA H200 | 1.71 | 35.19 | 0.002 | 689.8 W |
| NVIDIA H100 80GB HBM3 | 1.65 | 36.31 | 0.002 | 688.5 W |
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 1.13 | 53.1 | 0.002 | 569.8 W |
| NVIDIA A100 80GB SXM4 | 0.75 | 79.5 | 0.002 | 395.8 W |
What the numbers show. Across 4 GPUs measured on our own bench, H200 is fastest at 1.71 images/min. The slowest, A100 80GB SXM4, manages 0.75, so the spread is 2.3x from top to bottom. Per dollar of launch price, RTX PRO 6000 Blackwell Workstation Edition gives the most (0.1 images/min per $1,000).
How it compares. H100 80GB HBM3: Qwen-Image-Edit 2509 1.65 images/min, Qwen-Image 3.0 (20B), Qwen-Image 2512 3.0 (20B), Z-Image 3.58 (6B), Qwen-Image-Edit 3.68 (20B). All 4 beat Qwen-Image-Edit 2509 here.
Cost on a rented GPU. 1,000 images of Qwen-Image-Edit 2509: $15.87 on a RTX PRO 6000 Blackwell Workstation Edition ($1.08/hr, 14.7 hours), $34.99 on a H200 ($3.59/hr, 9.7 hours, 2.2x the cost).
Qwen-Image-Edit 2509: 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 | 1.13 | $15.87 |
| NVIDIA A100 80GB SXM4 | $0.95/hr | 0.75 | $21.04 |
| NVIDIA H100 80GB HBM3 | $2.14/hr | 1.65 | $21.58 |
| NVIDIA H200 | $3.59/hr | 1.71 | $34.99 |
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-Edit 2509. 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-Edit 2509 at 1024px and 40 steps: 35.1s per image on the H200, 80.0s on the A100 80GB SXM4. A batch of 100 takes 58 min on the fastest card and 133 min on the slowest.
VRAM for Qwen-Image-Edit 2509. Measured peak 59.4GB, so 80GB is the smallest common card size; smallest card it ran on: H100 80GB HBM3 (80GB).
Power on Qwen-Image-Edit 2509. Most efficient: H200, 690W, 6.72 kWh per 1,000 images. At $0.15/kWh: $1.01 per 1,000 images.
Fastest on Qwen-Image-Edit 2509: NVIDIA H200, 1.71 images/min. Cheapest to rent per job: NVIDIA RTX PRO 6000 Blackwell Workstation Edition, $15.87 per 1,000 images.
Qwen-Image-Edit 2509 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.