Alibaba's next-generation text-to-video model, released in public beta in August 2026. Wan 3.0 generates up to 30 seconds of coherent video at 1080p in a single pass, with continuous camera movement and one-take shot language. Its Omni-Reference feature accepts simultaneous multimodal references — documents, spreadsheets, presentations, webpages and markdown alongside text, images and audio — enabling direct document-to-video workflows from doc, ppt, pdf and xls inputs.
Parameters
Undisclosed
Context Window
License
Proprietary
Release Date
2026-08-07
Benchmark Performance
AA Intelligence Index
—
LMArena Elo
—
HLE
—
ARC-AGI-2
—
SWE-bench Verified
—
GPQA Diamond
—
MMLU-Pro
—
LiveCodeBench
—
AIME 2025
—
MATH-500
—
API Pricing
API pricing for this model is not yet available
Strengths
- •30-second single-pass generation at 1080p
- •Omni-Reference multimodal conditioning (doc/ppt/pdf/xls/text/image/audio)
- •Coherent camera motion and shot continuity without stitching
Weaknesses
- •Priced per second rather than per token ($0.05/s at 480p to $0.20/s at 1080p)
- •Public beta — capabilities and availability still stabilizing
- •Video-only modality with no text/instruct benchmark parity
Use Cases
- •Marketing and social video from documents or slides
- •Storyboarding and previsualization
- •Product and explainer videos from reference assets