DALL·E 3 size & quality picker
DALL·E 3 only offers three sizes and two quality levels, but the six resulting combinations have noticeably different prices. This picker shows the cost per image for any combination and multiplies it by your batch size so you can budget a generation run before you spend a cent.
How DALL·E 3 pricing works
OpenAI prices DALL·E 3 per generated image, and the price depends on two factors:
- Quality —
standardorhd. HD adds a second refinement pass for finer detail and roughly doubles the cost. - Size — the square
1024×1024is the cheapest; the wider1792×1024and taller1024×1792cost more because they contain more pixels.
There is no per-token billing here as with text models — it’s a flat fee per image, so total cost is simply price-per-image × number of images.
Tips
- Draft in standard, finish in HD. Iterate on composition cheaply at standard quality, then regenerate the winner in HD.
- Match aspect ratio to use. Landscape for hero banners, portrait for phone wallpapers and stories, square for avatars and thumbnails.
- Estimate before batching. A 200-image HD landscape run is materially more expensive than the same run at standard square — check the total here first.
The six combinations and their use cases
DALL-E 3 supports exactly three sizes and two quality levels, giving six possible combinations. Each suits different output contexts:
| Size | Quality | Best for |
|---|---|---|
| 1024×1024 | Standard | Avatars, thumbnails, social posts, rapid prototyping |
| 1024×1024 | HD | Product imagery, icons needing fine detail |
| 1792×1024 | Standard | Hero banners, website headers, YouTube thumbnails |
| 1792×1024 | HD | Print-destined landscape shots, high-res marketing |
| 1024×1792 | Standard | Mobile wallpapers, Instagram stories, portrait posts |
| 1024×1792 | HD | Print portraits, detailed tall compositions |
HD quality adds a refinement pass that improves fine texture, edges, and clarity on complex scenes. For simple or abstract imagery the difference between standard and HD is minimal, making HD most cost-effective when detail genuinely matters in the final use.
Why DALL-E 3 has only three sizes
Unlike text models that accept arbitrary input lengths, image diffusion models generate at fixed resolutions natively. DALL-E 3’s three sizes correspond to approximate aspect ratios (1:1, 16:9, 9:16) that cover most practical use cases. Arbitrary resolution would require post-generation upscaling or cropping, which changes the composition. For images needing a non-standard crop, generate at the closest native size and crop afterward.
Batch cost example
For example, 50 images at 1792×1024 HD costs noticeably more than 50 images at 1024×1024 standard. Using this picker, you can compare the two totals side by side before committing to a generation run — which often leads to choosing standard quality for the exploratory phase and only upgrading to HD for the final selections.