Table of Contents
Quick Verdict
SD 1.5 generates 512×512 images at 4.4 req/s. SDXL generates 1024×1024 images at 2.0 req/s. That 2.2x throughput advantage comes with a critical caveat: SD 1.5 produces images at one-quarter the resolution. On a dedicated GPU server, the comparison is not apples-to-apples — it is speed-at-512 versus quality-at-1024. Your API’s output requirements determine the winner.
For thumbnail generation, social media previews, and rapid prototyping, SD 1.5 is unbeatable on throughput. For production imagery that users will inspect at full resolution, SDXL’s 4x pixel count justifies the 2x latency cost.
Full data below. More at the GPU comparisons hub.
Specs Comparison
SDXL uses 4x more parameters than SD 1.5 (3.5B versus 860M) to achieve its quality improvement. The 75% higher VRAM footprint is the infrastructure cost of that upgrade.
| Specification | SD 1.5 | SDXL |
|---|---|---|
| Parameters | 860M (UNet) | 3.5B (UNet) |
| Architecture | Latent Diffusion | Latent Diffusion |
| Context Length | 512×512 | 1024×1024 |
| VRAM (FP16) | 4 GB | 7 GB |
| VRAM (INT4) | N/A | N/A |
| Licence | CreativeML Open RAIL-M | CreativeML Open RAIL++-M |
Guides: SD 1.5 VRAM requirements and SDXL VRAM requirements.
API Throughput Benchmark
Tested on an NVIDIA RTX 3090 at default resolutions. See our benchmark page.
| Model (INT4) | Requests/sec | p50 Latency (ms) | p99 Latency (ms) | VRAM Used |
|---|---|---|---|---|
| SD 1.5 | 4.4 | 401 | 3422 | 4 GB |
| SDXL | 2.0 | 711 | 2568 | 7 GB |
SDXL’s tighter p99 (2,568 ms versus 3,422 ms) despite higher median latency suggests more predictable generation times under load. SD 1.5’s wider p99 spread indicates some requests experience significant queueing. See our best GPU for LLM inference guide.
See also: SD 1.5 vs SDXL for Cost-Optimised Batch Processing for a related comparison.
See also: SDXL vs Flux.1 for API Serving (Throughput) for a related comparison.
Cost Analysis
SD 1.5’s cost advantage is roughly 46% per 1,000 images, but remember you are comparing 512×512 to 1024×1024 output.
| Cost Factor | SD 1.5 | SDXL |
|---|---|---|
| GPU Required | RTX 3090 (24 GB) | RTX 3090 (24 GB) |
| VRAM Used | 4 GB | 7 GB |
| Images/min | 4.1 | 2.2 |
| Cost/1K Images | £1.25 | £2.33 |
See our cost calculator.
Recommendation
Choose SD 1.5 for APIs generating thumbnails, avatars, or preview images where 512×512 resolution is acceptable and throughput determines infrastructure cost.
Choose SDXL for APIs delivering production-quality 1024×1024 images where resolution and detail directly affect the user experience — product imagery, marketing content, and design tool integrations.
Serve on dedicated GPU servers for consistent image generation throughput.
Deploy the Winner
Run SD 1.5 or SDXL on bare-metal GPU servers with full root access, no shared resources, and no token limits.
Browse GPU Servers
