Turning a prompt into a picture, timed end to end at 1024px.
One prompt, one seed, 1024px, rendered by each model on the same card (NVIDIA L4). Speed is meaningless without knowing what you get for it.




Prompt: "a photorealistic red fox standing on a mossy rock in a misty forest, golden hour". Identical seed and resolution for every model, at each model's own recommended step count, so a four-step model is not judged at thirty.
Weights: stabilityai/stable-diffusion-xl-base-1.0
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA B200 | 46.12 |
| 2 | NVIDIA H200 | 37.16 |
| 3 | NVIDIA H100 80GB HBM3 | 34.58 |
| 4 | NVIDIA B300 | 29.2 |
| 5 | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 27.94 |
| 6 | NVIDIA RTX PRO 6000 Blackwell Server Edition | 26.32 |
| 7 | NVIDIA GeForce RTX 5090 | 21.08 |
| 8 | NVIDIA A100 40GB SXM4 | 18.58 |
| 9 | NVIDIA RTX PRO 5000 Blackwell | 17.46 |
| 10 | NVIDIA L40S | 17.02 |
| 11 | NVIDIA A100 80GB PCIe | 16.8 |
| 12 | NVIDIA A100 80GB SXM4 | 16.36 |
| 13 | NVIDIA GeForce RTX 4090 | 16.28 |
| 14 | NVIDIA RTX PRO 4500 Blackwell | 13.1 |
| 15 | NVIDIA RTX 5000 Ada Generation | 12.74 |
Weights: Tongyi-MAI/Z-Image-Turbo
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA B300 | 39.68 |
| 2 | NVIDIA B200 | 32.33 |
| 3 | NVIDIA H200 | 23.18 |
| 4 | NVIDIA H100 80GB HBM3 | 22.13 |
| 5 | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 15.6 |
| 6 | NVIDIA RTX PRO 6000 Blackwell Server Edition | 14.78 |
| 7 | NVIDIA GeForce RTX 5090 | 10.73 |
| 8 | NVIDIA A100 80GB SXM4 | 9.98 |
| 9 | NVIDIA A100 80GB PCIe | 9.3 |
| 10 | NVIDIA RTX PRO 5000 Blackwell | 9.08 |
| 11 | NVIDIA L40S | 8.33 |
| 12 | NVIDIA GeForce RTX 4090 | 7.24 |
| 13 | NVIDIA RTX PRO 4500 Blackwell | 6.45 |
| 14 | NVIDIA RTX 5000 Ada Generation | 5.93 |
| 15 | NVIDIA L40 | 5.63 |
Weights: stable-diffusion-v1-5/stable-diffusion-v1-5
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA GeForce RTX 5090 | 102.36 |
| 2 | NVIDIA H100 80GB HBM3 | 80.45 |
| 3 | NVIDIA GeForce RTX 4090 | 80.14 |
| 4 | GeForce RTX 5080 | 61.19 |
| 5 | NVIDIA L40S | 58.54 |
| 6 | GeForce RTX 4080 Super | 51.96 |
| 7 | NVIDIA GeForce RTX 4080 | 51.23 |
| 8 | GeForce RTX 5070 Ti | 49.51 |
| 9 | NVIDIA GeForce RTX 4070 Ti Super | 46.09 |
| 10 | NVIDIA GeForce RTX 4070 Ti | 44.44 |
| 11 | NVIDIA GeForce RTX 3090 Ti | 43.77 |
| 12 | NVIDIA RTX PRO 4000 Blackwell | 40.91 |
| 13 | NVIDIA GeForce RTX 3080 Ti | 40.7 |
| 14 | NVIDIA GeForce RTX 4070 Super | 39.83 |
| 15 | NVIDIA RTX 4000 (Ada Generation) | 39.51 |
Weights: Efficient-Large-Model/Sana_1600M_1024px_diffusers
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA GeForce RTX 5090 | 51.52 |
| 2 | NVIDIA L40S | 50.81 |
| 3 | NVIDIA GeForce RTX 4090 | 41.47 |
| 4 | GeForce RTX 5080 | 30.02 |
| 5 | NVIDIA GeForce RTX 4080 | 27.92 |
| 6 | GeForce RTX 4080 Super | 26.98 |
| 7 | GeForce RTX 5070 Ti | 26.18 |
| 8 | NVIDIA GeForce RTX 4070 Ti Super | 22.59 |
| 9 | NVIDIA GeForce RTX 3090 Ti | 21.84 |
| 10 | NVIDIA GeForce RTX 4070 Ti | 21.64 |
| 11 | NVIDIA GeForce RTX 3090 | 20.6 |
| 12 | NVIDIA GeForce RTX 3080 Ti | 20.44 |
| 13 | NVIDIA RTX PRO 4000 Blackwell | 20.12 |
| 14 | GeForce RTX 5070 | 19.71 |
| 15 | NVIDIA GeForce RTX 4070 Super | 19.35 |
Weights: playgroundai/playground-v2.5-1024px-aesthetic
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA H100 80GB HBM3 | 21.48 |
| 2 | NVIDIA GeForce RTX 5090 | 13.28 |
| 3 | NVIDIA L40S | 10.85 |
| 4 | NVIDIA GeForce RTX 4090 | 9.96 |
| 5 | GeForce RTX 5080 | 7.6 |
| 6 | GeForce RTX 4080 Super | 6.74 |
| 7 | NVIDIA GeForce RTX 4080 | 6.33 |
| 8 | GeForce RTX 5070 Ti | 5.88 |
| 9 | NVIDIA GeForce RTX 4070 Ti Super | 5.57 |
| 10 | NVIDIA RTX PRO 4000 Blackwell | 5.54 |
| 11 | NVIDIA GeForce RTX 3090 Ti | 5.39 |
| 12 | NVIDIA GeForce RTX 4070 Ti | 5.29 |
| 13 | NVIDIA GeForce RTX 3080 Ti | 4.9 |
| 14 | NVIDIA GeForce RTX 3090 | 4.79 |
| 15 | NVIDIA GeForce RTX 4070 Super | 4.79 |
Weights: stabilityai/sdxl-turbo
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA GeForce RTX 5090 | 902.04 |
| 2 | NVIDIA GeForce RTX 4090 | 652.11 |
| 3 | NVIDIA H100 80GB HBM3 | 618.27 |
| 4 | NVIDIA GeForce RTX 4070 Ti Super | 532.27 |
| 5 | GeForce RTX 5080 | 512.76 |
| 6 | NVIDIA GeForce RTX 4080 | 484.82 |
| 7 | NVIDIA GeForce RTX 3090 Ti | 456.11 |
| 8 | NVIDIA GeForce RTX 3080 Ti | 440.89 |
| 9 | GeForce RTX 5070 Ti | 431.35 |
| 10 | NVIDIA L40S | 423.51 |
| 11 | GeForce RTX 4080 Super | 410.08 |
| 12 | GeForce RTX 5070 | 383.37 |
| 13 | NVIDIA GeForce RTX 3090 | 332.99 |
| 14 | NVIDIA GeForce RTX 4070 | 286.04 |
| 15 | GeForce RTX 5060 Ti | 260.84 |
Weights: black-forest-labs/FLUX.1-dev
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA B300 | 20.81 |
| 2 | NVIDIA B200 | 13.14 |
| 3 | NVIDIA H200 | 9.51 |
| 4 | NVIDIA H100 80GB HBM3 | 9.06 |
| 5 | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 6.92 |
| 6 | NVIDIA RTX PRO 6000 Blackwell Server Edition | 6.75 |
| 7 | NVIDIA A100 80GB SXM4 | 4.26 |
| 8 | NVIDIA RTX PRO 5000 Blackwell | 4.2 |
| 9 | NVIDIA A100 80GB PCIe | 3.96 |
| 10 | NVIDIA L40S | 3.84 |
| 11 | NVIDIA L40 | 2.46 |
| 12 | NVIDIA RTX 6000 Ada Generation | 2.23 |
| 13 | NVIDIA RTX A6000 | 2.16 |
| 14 | NVIDIA GeForce RTX 5090 | 2.04 |
| 15 | NVIDIA A40 | 1.95 |
Weights: PixArt-alpha/PixArt-Sigma-XL-2-1024-MS
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA GeForce RTX 5090 | 30.01 |
| 2 | NVIDIA L40S | 26.95 |
| 3 | NVIDIA GeForce RTX 4090 | 24.27 |
| 4 | GeForce RTX 5080 | 16.22 |
| 5 | NVIDIA GeForce RTX 4080 | 14.6 |
| 6 | GeForce RTX 4080 Super | 14.59 |
| 7 | NVIDIA RTX PRO 4000 Blackwell | 13.44 |
| 8 | GeForce RTX 5070 Ti | 13.34 |
| 9 | NVIDIA GeForce RTX 4070 Ti Super | 12.6 |
| 10 | NVIDIA GeForce RTX 3090 Ti | 11.56 |
| 11 | NVIDIA GeForce RTX 3090 | 10.9 |
| 12 | NVIDIA RTX 4000 (Ada Generation) | 10.45 |
| 13 | NVIDIA A10G | 9.93 |
| 14 | NVIDIA L4 | 8.08 |
| 15 | NVIDIA RTX A4000 | 8.05 |
Weights: Lykon/dreamshaper-xl-v2-turbo
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA GeForce RTX 5090 | 70.72 |
| 2 | NVIDIA GeForce RTX 4090 | 54.35 |
| 3 | NVIDIA GeForce RTX 4070 Ti Super | 30.49 |
| 4 | NVIDIA GeForce RTX 3080 Ti | 28.09 |
| 5 | NVIDIA GeForce RTX 3090 | 27.3 |
| 6 | NVIDIA GeForce RTX 4070 Super | 26.34 |
| 7 | NVIDIA GeForce RTX 4070 | 22.67 |
| 8 | GeForce RTX 5060 Ti | 18.31 |
| 9 | NVIDIA GeForce RTX 4060 Ti 16GB | 16.44 |
| 10 | NVIDIA GeForce RTX 3060 | 10.14 |
Weights: Tongyi-MAI/Z-Image-Turbo
Measured in 1024px. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA B200 | 62.47 |
| 2 | NVIDIA B300 | 48.63 |
| 3 | NVIDIA H200 | 40.74 |
| 4 | NVIDIA H100 80GB HBM3 | 39.42 |
| 5 | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 31.43 |
| 6 | NVIDIA A100 40GB SXM4 | 18.7 |
| 7 | NVIDIA A100 80GB SXM4 | 18.65 |
| 8 | NVIDIA L40S | 18.2 |
| 9 | NVIDIA A10G | 6.32 |
| 10 | NVIDIA L4 | 4.6 |
Weights: SimianLuo/LCM_Dreamshaper_v7
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA GeForce RTX 4090 | 260.94 |
| 2 | GeForce RTX 5070 Ti | 179.8 |
| 3 | GeForce RTX 5070 | 142.36 |
| 4 | GeForce RTX 5060 Ti | 106.19 |
| 5 | NVIDIA GeForce RTX 4060 Ti 16GB | 94.29 |
| 6 | NVIDIA GeForce RTX 3070 Founders Edition | 92.56 |
| 7 | NVIDIA GeForce RTX 3060 Ti | 80.24 |
| 8 | NVIDIA GeForce RTX 3060 | 62.2 |
Weights: black-forest-labs/FLUX.1-schnell
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA B200 | 81.41 |
| 2 | NVIDIA H200 | 60.51 |
| 3 | NVIDIA B300 | 59.54 |
| 4 | NVIDIA H100 80GB HBM3 | 58.36 |
| 5 | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 42.41 |
| 6 | NVIDIA A100 40GB SXM4 | 28.44 |
| 7 | NVIDIA A100 80GB SXM4 | 28.39 |
| 8 | NVIDIA L40S | 26.24 |
Weights: stabilityai/sd-turbo
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA GeForce RTX 4090 | 786.15 |
| 2 | GeForce RTX 5070 Ti | 644.31 |
| 3 | GeForce RTX 5070 | 535.06 |
| 4 | NVIDIA GeForce RTX 4060 Ti 16GB | 361.33 |
| 5 | NVIDIA GeForce RTX 3070 Founders Edition | 357.81 |
| 6 | NVIDIA GeForce RTX 3060 Ti | 327.8 |
| 7 | NVIDIA GeForce RTX 3060 | 267.93 |
Weights: segmind/SSD-1B
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA GeForce RTX 5090 | 20.49 |
| 2 | NVIDIA GeForce RTX 4090 | 15.77 |
| 3 | NVIDIA GeForce RTX 3090 | 7.48 |
| 4 | GeForce RTX 5060 Ti | 4.56 |
| 5 | NVIDIA GeForce RTX 4060 Ti 16GB | 4.49 |
| 6 | NVIDIA GeForce RTX 3070 Founders Edition | 4.31 |
| 7 | NVIDIA GeForce RTX 3060 | 2.72 |
Weights: fal/AuraFlow-v0.3
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA GeForce RTX 5090 | 4.35 |
| 2 | NVIDIA L40S | 3.58 |
| 3 | NVIDIA GeForce RTX 4090 | 3.02 |
| 4 | NVIDIA GeForce RTX 3090 Ti | 1.67 |
| 5 | NVIDIA GeForce RTX 3090 | 1.5 |
| 6 | NVIDIA A10G | 1.39 |
| 7 | NVIDIA L4 | 1.02 |
Weights: Tongyi-MAI/Z-Image
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA H100 80GB HBM3 | 3.58 |
| 2 | NVIDIA GeForce RTX 5090 | 1.72 |
| 3 | NVIDIA L40S | 1.31 |
| 4 | NVIDIA GeForce RTX 4090 | 1.17 |
| 5 | NVIDIA GeForce RTX 3090 Ti | 0.7 |
| 6 | NVIDIA GeForce RTX 3090 | 0.63 |
| 7 | NVIDIA L4 | 0.37 |
Weights: Lykon/dreamshaper-xl-lightning
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA GeForce RTX 5090 | 122.85 |
| 2 | NVIDIA GeForce RTX 4090 | 91.52 |
| 3 | NVIDIA GeForce RTX 3090 | 47.54 |
| 4 | GeForce RTX 5060 Ti | 28.35 |
| 5 | NVIDIA GeForce RTX 4060 Ti 16GB | 27.95 |
| 6 | NVIDIA GeForce RTX 3060 | 18.28 |
Weights: black-forest-labs/FLUX.2-klein-4B
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA H100 80GB HBM3 | 108.33 |
| 2 | NVIDIA GeForce RTX 5090 | 59.33 |
| 3 | NVIDIA L40S | 50.29 |
| 4 | NVIDIA GeForce RTX 4090 | 42.38 |
| 5 | NVIDIA GeForce RTX 3090 Ti | 23.68 |
| 6 | NVIDIA GeForce RTX 3090 | 21.26 |
Weights: krea/Krea-2-Turbo
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA B200 | 9.5 |
| 2 | NVIDIA H200 | 7.25 |
| 3 | NVIDIA H100 80GB HBM3 | 6.85 |
| 4 | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 4.71 |
| 5 | NVIDIA A100 80GB SXM4 | 3.35 |
Weights: Manojb/stable-diffusion-2-1-base
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA GeForce RTX 5090 | 65.53 |
| 2 | NVIDIA GeForce RTX 4090 | 60.58 |
| 3 | NVIDIA GeForce RTX 4060 Ti 16GB | 28.74 |
| 4 | NVIDIA GeForce RTX 3060 | 17.71 |
Weights: baidu/ERNIE-Image-Turbo
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA A100 80GB SXM4 | 10.67 |
| 2 | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 7.47 |
| 3 | NVIDIA H200 | 5.52 |
| 4 | NVIDIA L40S | 4.41 |
Weights: Qwen/Qwen-Image
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA H200 | 3.14 |
| 2 | NVIDIA H100 80GB HBM3 | 3 |
| 3 | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 1.97 |
| 4 | NVIDIA A100 80GB SXM4 | 1.34 |
Weights: Qwen/Qwen-Image-2.1
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA H100 80GB HBM3 | 4.59 |
| 2 | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 3.21 |
| 3 | NVIDIA A100 80GB SXM4 | 2.06 |
| 4 | NVIDIA L40S | 1.78 |
Weights: Kwai-Kolors/Kolors-diffusers
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA GeForce RTX 5090 | 13.53 |
| 2 | NVIDIA GeForce RTX 4090 | 9.88 |
| 3 | NVIDIA GeForce RTX 3090 | 4.69 |
Weights: lodestones/Chroma1-HD
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA GeForce RTX 5090 | 1.29 |
| 2 | NVIDIA A100 40GB PCIe | 1.07 |
| 3 | NVIDIA RTX A6000 | 0.62 |
Weights: stabilityai/stable-diffusion-3.5-medium
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA L40S | 11.58 |
| 2 | NVIDIA A10G | 4.11 |
| 3 | NVIDIA L4 | 2.96 |
Weights: Qwen/Qwen-Image-2512
Measured in images/min. Longer is faster.
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA H200 | 3.11 |
| 2 | NVIDIA H100 80GB HBM3 | 3 |
| 3 | NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 1.96 |
Weights: black-forest-labs/FLUX.2-klein-9B
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA H100 80GB HBM3 | 64.87 |
| 2 | NVIDIA L40S | 25.95 |
Weights: stabilityai/stable-diffusion-3-medium-diffusers
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA GeForce RTX 4090 | 13.95 |
Weights: ByteDance/SDXL-Lightning
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA GeForce RTX 4090 | 86.71 |
Weights: HiDream-ai/HiDream-I1-Fast
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA H200 | 14.9 |
Weights: stabilityai/stable-diffusion-3.5-large
| # | GPU | images/min |
|---|---|---|
| 1 | NVIDIA L40S | 4.54 |
Every number is a first-party run on our own bench. Cards missing from a board have not been measured on that model yet, or cannot fit it.