Alibaba released Wan 2.7 Image on April 1, 2026 — and despite the date, it is not a joke. The Wan 2.7 release is image-first, not video. The Wan video stack remains Wan 2.6. What launched is a unified image generation and editing model that introduces reasoning to the generation process, raises the output ceiling to 4K, and significantly advances what is possible with multilingual text-in-image content.
Two model variants launched simultaneously: Wan2.7-Image for standard generation and editing, and Wan2.7-Image-Pro for more stable compositions, sharper prompt understanding, and 4K native output.
Thinking Mode: Reasoning Before Generating
The most architecturally significant addition in Wan 2.7 is chain-of-thought reasoning applied to image generation. Most image models process a prompt in a single forward pass — fast, but with limited ability to handle complex compositional logic, precise spatial relationships, or multi-element scenes where elements need to relate to each other correctly.
Wan 2.7's thinking mode adds a reasoning step before the generation begins. The model analyzes the prompt's spatial relationships, composition logic, and semantic intent — then generates based on that analysis. The practical result: images that match complex instructions more accurately, text that is correctly rendered, and edits that preserve intended context.
This reasoning step takes time. For prompts where you need a specific composition with multiple precisely positioned elements, it is worth the extra seconds. For simple single-subject generations, standard mode is faster.
What Changed From Wan 2.6
Architecture. Wan 2.7 maps text and visual semantics into a shared latent space. The previous generation used separate processing streams that connected at generation time. The unified space means the model has underlying semantic understanding of both domains simultaneously — it is not guessing at visual intent from text description, it is reasoning within a shared representation.
Text rendering. Up to 3,000 tokens of text input, support for 12 languages, and print-quality rendering of structured content. Wan 2.7 can generate infographics, tables, academic formulas, A4-formatted documents, and multi-language poster layouts with accurate text at production quality. For any workflow that requires readable text within generated images — marketing materials, bilingual content, structured visual layouts — this is a meaningful capability upgrade.
Reference capacity. Up to 9 reference images for personalization and consistency control. Outputs up to 12 images in a single generation — useful for storyboard packs, style variation sets, and product-scene iteration at volume.
Color control. A new color palette input system: specify exact color codes and proportions in the prompt, and the model matches them precisely. For brand-accurate asset generation — company colors, campaign palettes, product color variants — this eliminates the iterative correction that brand work typically requires.
4K output. Pro variant generates natively at 4K rather than upscaling from a lower resolution. This preserves fine texture detail, edge sharpness, and font legibility that upscaling consistently softens.
Wan 2.7 Is Not a Video Release
This is worth stating clearly because "Wan 2.7" as a search term conflates the image launch with expectations of a video model.
Alibaba's official material for Wan 2.7 is entirely image-focused. The current publicly documented and API-available Alibaba video generation stack is Wan 2.6 — with wan2.6-t2v and wan2.6-i2v model families, public per-second pricing, and multi-shot narrative support up to 15 seconds.
Wan 2.7 image capabilities are strong. If your workflow starts with image generation and ends with an image, Wan 2.7 is the right Alibaba model to evaluate. If your workflow ends with a video clip, Wan 2.6 is still the Alibaba model to use. They are not competing for the same job.
Availability
Wan 2.7 Image and Wan 2.7 Image Pro are available through Alibaba's Model Studio platform, the official Wan website (wan.video), and are being integrated into the Qwen App. API access is available through Model Studio with standard Alibaba Cloud pricing.
