Wan 3 vs Pro vs Prime: a controlled 9-run benchmark
Three models, one prompt, three runs each. We report completion time, live 1080P price, technical success, and what the frames actually show.First-party runRun September 20, 2026Registry 2026-09-22.3
What the test found
All nine jobs completed and returned a 5.04-second 1080P H.264 video with AAC audio. Median completion times were within 1.3 seconds of one another, so this sample does not support a meaningful speed winner.| Model | Completed | Median time | Live 5s price |
|---|---|---|---|
| Wan 3 Spicy | 3/3 | 150.0 s | $0.77 |
| Wan 3 Pro Spicy | 3/3 | 149.0 s | $0.693 |
| Wan 3 Prime Spicy | 3/3 | 148.7 s | $1.078 |
Method
Text-to-video, 16:9, 1080P, 5 seconds, audio enabled. Each model received the identical prompt three times. Timings are submit-to-complete wall time. This is a small controlled sample, not a universal quality ranking.Prompt usedA premium cobalt-blue ceramic perfume bottle on a pale stone plinth in a quiet studio. One continuous five-second shot: the camera makes a slow clockwise orbit while a narrow warm light sweeps across the bottle. Keep the bottle shape, label placement, plinth, and background stable. Natural reflections, restrained motion, no cuts, no extra objects, no text changes.
Visual review
Every run kept one blue bottle, one plinth, and a continuous studio shot. Shape and label styling varied between independent runs, as expected from text-to-video. Within each clip, the subject remained coherent in the sampled early, middle, and late frames. The 3×3 sheets below show those samples; each row is one run.Contact sheets
Wan 3 Spicy
3 runs · modelrush/wan3-spicy
Wan 3 Pro Spicy
3 runs · modelrush/wan3-pro-spicy
Wan 3 Prime Spicy
3 runs · modelrush/wan3-prime-spicy
ModelRushOne integration, intelligent routing, transparent billing. Model infrastructure for developers and agents.© 2026 ModelRushAll systems operational