Resize AI 2026 Performance —
Great Things Come to Those Who Wait…
ON1 Resize AI 2026 introduces two cutting-edge machine learning models designed to deliver the highest quality photo enlargements yet.
The first, the Standard model, is a direct upgrade over Resize AI 2023. It delivers noticeably better results—particularly with smaller, lower-quality, or noisier source images. While it takes slightly longer to process than the 2023 model, it remains fast enough for batch processing and uses fewer GPU resources, making it ideal for systems with older or less powerful graphics cards.
The second model, Highest Quality, represents a major leap forward. Built on diffusion-based generative AI, it produces stunning, lifelike detail that goes beyond the original photo. It can reduce noise and even synthesize realistic details, making it feel like you're shooting with a 4× teleconverter or a camera with four times the megapixels. This model is perfect for making dramatic crops from noisy or low-res photos and still achieving breathtaking enlargements.
However, this enhanced quality comes at a cost—it’s more computationally intensive and highly GPU-dependent. Performance scales with GPU power: the newer your GPU, the faster the results.
Below is a real-world example. We cropped a small 1100×1400 section from a 5500×3700 image and enlarged it 4× to 4500×5600. Here’s how both models performed across a range of GPUs.
Photo: Brett Sayles
Timings with Various Resize AI Models & GPUs
| Apple M1 (2020) | Apple M3 Max (2023) | NVIDIA GTX 1070 (2016) | NVIDIA RTX 4060 (2023) | AMD Radeon 5600 (2019) | AMD Radeon 7900XT (2022) | |
| Standard Model | 0:33 | 0:11 | 0:22 | 0:08 | 0:05 | 0:04 |
| Highest Quality Model | 1:34 | 0:35 | 1:09 | 0:23 | 0:40 | 0:12 |
Shorter Times are Better
As you can see, the more modern the GPU, generally with more dedicated VRAM and AI processing cores, has a huge impact on processing times. This is true with any of the AI features in other ON1 products as well. If you are looking to upgrade your computer, spending on the GPU is your best investment.
Notes: The NVIDIA 1xxx series and earlier are no longer supported by NVIDIA. AMD Radeon 5xx and earlier are no longer supported by AMD. These will be painfully slow and may stop working with future updates. Intel integrated GPUs are not supported.
Comments
0 comments
Article is closed for comments.