que paso
Dataconomy reports that Alibaba launched Wan3.0 on August 24, expanding a model that had been in public beta since August 6. The report says Wan3.0 can generate videos of up to 30 seconds in a single pass, twice the reported limit of Wan2.7, and supports resolutions up to 1080p.
Dataconomy reports that Alibaba launched Wan3.0 on August 24, 2026, after making the model available in public beta on August 6 through Alibaba Cloud’s Model Studio and Qwen Cloud. The article describes the launch as an expansion of access to Alibaba’s latest AI video-generation system. The visible source date places the development inside the current 96-hour news window. The report also links the timing to Alibaba’s $10.2 billion share placement in Hong Kong, saying proceeds were earmarked for artificial-intelligence investment. That financing context is reported by Dataconomy and is not independently confirmed here as a direct funding source for Wan3.0 itself.
According to Dataconomy, Wan3.0 can create clips of up to 30 seconds in a single pass, compared with a reported 15-second limit for Wan2.7. The article says the system supports output resolutions up to 1080p and is intended to preserve character detail, spatial layout, and motion consistency throughout a clip. It also reports features for synchronized facial micro-expressions and multilingual voice output. These are product capabilities described in the source; Dataconomy does not provide independent benchmark results, test procedures, sample-level evaluations, or comparisons demonstrating that the model meets those claims across different prompts and subjects.
Dataconomy says Wan3.0 accepts text, images, audio, video, web pages, and documents including PDFs and PowerPoint files as inputs. The article presents this as a way to turn structured material such as slide decks and spreadsheets into video, with particular relevance for marketing and corporate-communications teams. The source does not specify which document formats are supported in practice beyond its examples, how the system handles copyrighted or private material, or whether generated audio and visuals receive provenance labels.
The report lists pricing of $0.05 per second for 480p output, $0.10 per second for 720p, and $0.20 per second for 1080p. At the highest listed rate, Dataconomy calculates a cost of $12 per minute. The article compares the 1080p rate with a reported $0.40-per-second standard-service price for Google’s Veo 3.1. Dataconomy says access remains controlled through an application process on Model Studio and Qwen Cloud, while a consumer-facing site at wan.video is planned as a members-only platform. The report also says Alibaba has closed Wan3.0’s model weights, marking a departure from the open-source approach associated with earlier Wan releases.
Lea la fuente principal: dataconomy.com ↗
Por qué es importante
The reported release is a meaningful product change in AI video generation because it combines longer single-pass output, multiple input formats, and a published usage-price structure. However, the report does not provide independent testing of output quality, reliability, or safety.
The reported longer output limit matters because a 30-second generation can cover more of a short advertisement, explainer, product sequence, or social-video concept without requiring users to assemble as many separate clips. Fewer cuts may reduce editing work and continuity problems, although the source offers no independent evidence that Wan3.0 actually maintains continuity better than competing systems. Longer output alone does not establish better storytelling, factual accuracy, visual control, or suitability for professional production.
The combination of multimodal inputs and document-to-video generation could make the product useful to teams that already work from presentations, written briefs, or recorded material. If the reported workflow is reliable, it could shorten the path from existing communications material to a draft video. That practical value depends on permissions, data handling, editability, audio quality, and the ability to correct individual scenes. None of those operational details is established by the Dataconomy report.
The pricing information gives prospective users a concrete way to estimate generation costs and places Wan3.0 in direct competition with other commercial video models. Dataconomy describes the pricing as competitive, but the comparison is incomplete without information about retries, input fees, storage, audio charges, rate limits, commercial rights, and the amount of usable footage produced per attempt. The report’s financial context also highlights the scale of Alibaba’s AI investment: it cites a 45% year-over-year increase in cloud and AI revenue to 48.44 billion yuan for the June quarter, alongside a 75% decline in quarterly net profit and capital expenditure of 67.68 billion yuan. Those figures are reported context, not evidence that Wan3.0 will succeed commercially.
Qué ver a continuación
The main questions are whether Wan3.0 becomes broadly accessible, whether its reported character, motion, voice, and layout capabilities hold up in independent tests, and how Alibaba’s closed-weight strategy affects developers who previously relied on open Wan models.
Independent evaluation is the most important next step. Reviewers and researchers should test Wan3.0 across long scenes, changing camera movements, multiple characters, text rendered inside images, culturally varied prompts, and multilingual speech. They should distinguish between the model’s maximum technical output length and the proportion of generations that remain coherent and usable for the full 30 seconds. Dataconomy explicitly says the model has not yet been independently benchmarked, so Alibaba’s performance claims remain unverified in the supplied source.
Access and terms will determine whether the model is practically available. The report describes application-controlled access through Model Studio and Qwen Cloud and a planned members-only consumer site. It does not say how many users have been accepted, where the service is available, whether access is free or paid during the application period, or whether the listed prices apply uniformly across platforms. It also does not explain whether users can export, edit, or commercially publish generated material under the same terms.
The closed model weights create a significant change for developers and researchers who might otherwise run or inspect the system themselves. Watch for clarification about whether Alibaba will publish technical documentation, model cards, safety policies, content restrictions, training-data disclosures, or evaluation results. The source also leaves open questions about copyright, consent for uploaded documents and media, provenance marking, misuse prevention, and the handling of realistic faces and voices. Finally, Dataconomy reports that Alibaba had committed 380 billion yuan to AI infrastructure over three years and that reports suggested a possible increase to 480 billion yuan; neither the proposed increase nor its relationship to Wan3.0 is confirmed in the supplied material.
