Wan 3.0 Prime vs Standard Wan 3.0

Wan 3.0 Prime is the same Wan 3.0 API at 1.5x the price, and the premium buys speed rather than new controls. Every parameter, resolution, duration range, and aspect ratio is identical between alibaba/wan-3.0-prime/* and alibaba/wan-3.0/*. If your pipeline is latency-bound, Prime is worth it. If it is cost-bound, standard Wan 3.0 is the same model family with the same inputs at two thirds the price.

This post lays out both price ladders and a decision rule. All prices are computed from the live WaveSpeedAI catalog formulas on September 5, 2026.

What Prime is

Standard Wan 3.0 went live on WaveSpeedAI on August 3, 2026, with three endpoints: text-to-video, image-to-video, and reference-to-video. The Prime variants of all three followed on August 24.

Alibaba’s own documentation for the model describes it as follows: “Wan3.0-Video-Prime is the accelerated version of the Wan video generation model, delivering significantly faster generation speed while maintaining high-quality output” (Alibaba Cloud Model Studio). The same page lists the same capabilities as standard Wan 3.0: text, image, video, and audio inputs; 480p, 720p, and 1080p output; up to 30 seconds.

That page does not state a speed multiplier, and we have not measured one in a controlled comparison, so this post does not quote a number. Treat Prime as the faster queue for the same job.

	Standard Wan 3.0	Wan 3.0 Prime
text-to-video	alibaba/wan-3.0/text-to-video	alibaba/wan-3.0-prime/text-to-video
image-to-video	alibaba/wan-3.0/image-to-video	alibaba/wan-3.0-prime/image-to-video
reference-to-video	alibaba/wan-3.0/reference-to-video	alibaba/wan-3.0-prime/reference-to-video
Live since	August 3, 2026	August 24, 2026
Base price (720p, per 5s)	$0.50	$0.75

Parameters: nothing changes

The input schemas are field-for-field identical. This is the shared surface.

Parameter	text-to-video	image-to-video	reference-to-video
prompt	required	required	required
image	-	required (URL or base64)	-
last_image	-	optional last frame	-
reference_images	-	-	up to 10
reference_videos	-	-	up to 5, 1-15s each, 15s total
reference_audios	-	-	up to 5, 15s total
resolution	480p, 720p, 1080p (default 720p)	same	same
aspect_ratio	16:9, 9:16, 1:1, 4:3, 3:4 (default 16:9)	same, or adapts to the image	same
duration	2-30 (default 5)	2-30 (default 5)	2-30; input plus output must not exceed 30
enable_audio	default true	default true	default true
enable_prompt_expansion	default false	default false	default false
seed	-1 for random	-1 for random	-1 for random

Reference-to-video requires at least one of reference_images, reference_videos, or reference_audios, and references are addressed by their array order in the prompt.

Switching variants is a one-string change in the request path. The pattern is the same for every WaveSpeedAI model: POST https://api.wavespeed.ai/api/v3/<model_uuid>, then poll GET /api/v3/predictions/<id>/result.

# Standard
curl -X POST https://api.wavespeed.ai/api/v3/alibaba/wan-3.0/text-to-video \
  -H "Authorization: Bearer $WAVESPEED_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "A slow dolly through a rain-soaked night market, neon reflections", "resolution": "720p", "duration": 5}'

# Prime: only the model_uuid changes
curl -X POST https://api.wavespeed.ai/api/v3/alibaba/wan-3.0-prime/text-to-video \
  -H "Authorization: Bearer $WAVESPEED_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "A slow dolly through a rain-soaked night market, neon reflections", "resolution": "720p", "duration": 5}'

Price ladder

Both variants use the same formula shape. Base price is per 5 seconds at 720p; 480p halves it and 1080p doubles it. For text-to-video and image-to-video the billed duration is the requested duration, rounded up to a whole second and clamped to 2-30. Audio is included at no extra cost on both variants.

Per second

Resolution	Standard	Prime	Premium
480p	$0.05	$0.075	+$0.025 per second
720p	$0.10	$0.15	+$0.05 per second
1080p	$0.20	$0.30	+$0.10 per second

Per clip: text-to-video and image-to-video

Resolution	Duration	Standard	Prime
480p	5s	$0.25	$0.375
480p	10s	$0.50	$0.75
480p	30s	$1.50	$2.25
720p	5s	$0.50	$0.75
720p	10s	$1.00	$1.50
720p	30s	$3.00	$4.50
1080p	5s	$1.00	$1.50
1080p	10s	$2.00	$3.00
1080p	30s	$6.00	$9.00

Reference-to-video

Reference-to-video bills output duration plus reference video input. Each reference video is padded to at least 1 second, the combined input is capped at 15 seconds and rounded up to a whole second, and that input total is billed at the same per-second rate as the output. Reference images and audio add nothing.

Resolution	5s output, 5s reference video	Standard	Prime
480p	10 billed seconds	$0.50	$0.75
720p	10 billed seconds	$1.00	$1.50
1080p	10 billed seconds	$2.00	$3.00

With images only and no reference video, reference-to-video costs the same as text-to-video for the same output duration.

When Prime is worth 1.5x

The premium is flat, so the decision does not depend on resolution or duration. It depends on whether generation latency is on your critical path.

Choose Prime when:

• A user is waiting on the result in an interactive product, and turnaround time is a feature you charge for or are judged on.
• A batch has a hard wall-clock deadline, such as assets for a live event that must be ready before it starts.
• You are iterating on a prompt or reference set and the cost of your own time exceeds the extra $0.05 per 720p second.

Choose standard Wan 3.0 when:

• Generation is asynchronous and nobody is watching a spinner: nightly batches, catalog-wide product videos, webhook-driven pipelines.
• You are producing at volume. Over 1,000 five-second 720p clips the difference is $250.
• You are generating 30-second 1080p clips, where the absolute premium is largest at $3.00 per clip.

A workable split: draft on standard at 480p ($0.25 per 5 seconds), and reserve Prime for the final render only when a deadline is close. Because the schema is identical, the same request body works on both, so the switch is a config value rather than a code change.

Related reading

• Wan 3.0 480p vs 720p vs 1080p for choosing a resolution before choosing a variant.
• How much does Wan 3.0 cost? for the standard price ladder in more detail.
• Does Wan 3.0 support reference-to-video? for the reference workflow itself.

Model pages: Wan 3.0 Prime text-to-video and standard Wan 3.0 text-to-video. The full family is on the Wan 3.0 API page.