TechApplication
Alibaba ships Wan3.0 video model with 30-second single-shot output at half Google Veo's price
The Next Web · Dataconomy · Yahoo Finance · 2026-08-24 · Source ↗
Alibaba ships Wan3.0 video model with 30-second single-shot output at half Google Veo's price
The Next Web · Dataconomy · Yahoo Finance · 2026-08-24 · Source ↗
Alibaba Cloud shipped Wan3.0 to general beta users on Monday, August 24. The public beta had opened on August 6, and Monday's release upgrades the model to a $0.05, $0.10 and $0.20-per-second three-tier price sheet at 480p, 720p and 1080p respectively, with the highest tier equivalent to $12 per minute. The comparable Google Veo 3.1 standard tier sells at $0.40 per second, meaning Wan3.0's 1080p output undercuts Veo by half on raw generation cost before Alibaba's tuning discounts. Weights remain closed, breaking the Wan2.2 open-source posture and mirroring the same closed-weights strategy Alibaba used for Qwen3.8-Max in early August.
"Wan3.0 accepts documents, spreadsheets, slides, and web pages as inputs and can produce videos as long as 30 seconds." — Alibaba Cloud release, August 24, 2026
What Wan3.0 does that Wan2.7 could not
The 30-second single-shot ceiling is double Wan2.7's 15-second limit and doubles the practical addressable market for the model. Wan3.0 outputs at up to 1080p, preserves character identity across the full 30-second window, holds spatial-layout consistency, and generates synchronized micro-expressions along with multilingual voice output including Mandarin, English, Japanese and Korean. The input surface widens beyond text and images to include documents, spreadsheets, slides and web pages, which is the surface Alibaba is pitching to corporate marketing teams that want to feed a product one-pager or PowerPoint deck to the model and receive an ad-ready video without frame-level editing.
The $10.2 billion share placement and the AI-infra earmark
The launch runs 72 hours behind Alibaba's Hong Kong share placement, which closed the previous Wednesday at HK$77.5 billion (roughly $10.2 billion) and represented Asia's largest single-tranche follow-on placement of 2026. Alibaba Cloud's own capex plan through fiscal 2028 is $53 billion, and the placement proceeds were earmarked entirely for AI infrastructure. Wan3.0 is the first customer-visible artifact of the placement, and Alibaba's messaging on Monday paired the model launch with a preview of the Qwen-UI-Agent GUI operator, a robotics-oriented Qwen-VL-Vision update, and expansion of the Qwen Cloud consumer surface at wan.video.
"Generative AI competition is increasingly becoming a contest between companies." — TechNode, on the Wan3.0 launch
Access and rollout
Beta access is application-based through Alibaba Cloud's Model Studio and Qwen Cloud platform. Full API access rolls out gradually to enterprise customers with existing Model Studio contracts, and a consumer platform at wan.video is planned as members-only. Alibaba said the beta had already been used across short-drama and film production, advertising and marketing, tourism promotion and music-video creation. Independent benchmarking against Veo, Runway Gen-5, Pika 2.5 and Kuaishou Kling has not been published; Alibaba's own comparison materials focused on Wan2.7 baselines and Veo 3.1 price rather than head-to-head quality benchmarks.
The strategic framing
Wan3.0 is Alibaba's most direct move to compete with Google's Veo line on the video-generation surface, and the closed-weights posture is a deliberate choice to protect the pricing power the $0.05-to-$0.20 tier gives Alibaba Cloud on Chinese enterprise contracts. The open-weights posture of Wan2.2 had produced strong developer mindshare but limited commercial capture, and Alibaba's read of the Wan2.7 upgrade cycle appears to have been that the enterprise buyers responsible for the bulk of paying revenue value model quality and price more than open weights. The Qwen-UI-Agent and Qwen-VL-Vision announcements paired with the launch signal Alibaba is defending the same vertical stack across coding agents, GUI operators and video generation.