Choose your graphics card to see which AI image and video models it runs on your own PC — Stable Diffusion 1.5 and XL, Stable Diffusion 3.5, FLUX.1 and FLUX.2, Qwen-Image and Wan 2.2 — at which precision they fit, and about how long each image or video takes. As of 9 November 2026.
| Model | Runs? | Time |
|---|
The smallest graphics card each model runs on entirely, at the most compressed precision people use, and the time each takes on two popular cards at its usual settings.
| Model | Makes | Smallest card | RTX 3060 12 GB | RTX 4090 |
|---|---|---|---|---|
| Stable Diffusion 1.5 | a 512 × 512 image | 4 GB (FP16) | 3 s–5 s | under a second |
| Stable Diffusion XL | a 1024 × 1024 image | 6 GB (FP8) | 12 s–24 s | 2 s–4 s |
| Stable Diffusion 3.5 Medium | a 1024 × 1024 image | 6 GB (FP8) | 12 s–25 s | 2 s–4 s |
| Stable Diffusion 3.5 Large | a 1024 × 1024 image | 10 GB (GGUF Q6_K) | 40 s–80 s | 7 s–14 s |
| FLUX.1 schnell | a 1024 × 1024 image | 12 GB (GGUF Q5_K_M) | 10 s–20 s | 2 s–3 s |
| FLUX.1 dev | a 1024 × 1024 image | 12 GB (GGUF Q5_K_M) | 49 s–2 min | 8 s–15 s |
| FLUX.2 dev | a 1024 × 1024 image | 24 GB (GGUF Q4_K_M) | Won’t fit | 32 s–65 s |
| Qwen-Image | a 1024 × 1024 image | 16 GB (GGUF Q4_K_M) | 4 min–8 min | 19 s–38 s |
| Wan 2.2 5B (text and image to video) | a 5-second 720p video | 10 GB (GGUF Q5_K_M) | 35 min–71 min | 6 min–13 min |
| Wan 2.2 14B (A14B) | a 5-second 480p video | 16 GB (GGUF Q5_K_M) | 55 min–1.8 hours | 4 min–8 min |
ComfyUI runs every model here, with ready-made workflows for each. For GGUF files it needs the ComfyUI-GGUF add-on; the model pages say which files fit which cards. Each model’s page links its files on Hugging Face.
Image and video models are limited by the graphics card’s arithmetic rather than its memory speed, so each card’s speed here comes from its measured arithmetic in llama.cpp’s benchmark (the same figure the AI model calculator uses for reading prompts), and each model’s time per step is anchored to measured times: FLUX.1 dev at about 1.85 steps a second on an RTX 4090, SDXL at about 4.2 a second on an RTX 3090, and Wan 2.2 5B taking about nine minutes for a 720p clip on an RTX 4090, its makers’ own figure. Other models scale from those by size. Memory is the model at each precision plus its working memory at the usual resolution, with about 0.8 GB kept for the driver and the screen. Speeds are for NVIDIA cards; AMD and Intel cards run these models through ROCm, ZLUDA, DirectML or Intel’s own builds, whose speed varies too much with the software to estimate. A real time is usually within the range shown; speed-up add-ons such as fewer-step LoRAs, TensorRT or SageAttention make it faster.
Bottleneck’s free app reads your graphics card, how much of its memory other programs are holding, your memory’s speed and channels, and the settings that slow a PC down, and says how to fix each. Its AI models screen lists the chat models your PC runs, and checks one you already run with Ollama or LM Studio.
The free version gives each check’s verdict, with no time limit and no account. Pro, $9 once, explains the cause and the fix. Compare the versions