Pricing
Blog

Sumi 2.6: new models and a refreshed workspace

· Sumi Studio Team

Sumi Studio 2.6 is live. It adds GPT Image 2.6, Seedream 5.5, Seedance 2.5 Flash and Qwen 4 Max, brings 4K to Seedance 2.5, and lets you edit an image by drawing on it. The home page and the Sumi window have been redesigned too. Here's what changed.

GPT Image 2.6: Flare and Sunburst

GPT Image 2.6 comes in two versions. Flare is the fast one, and OpenAI says it handles layouts like posters and product shots better. Sunburst is for detailed work and keeps details more consistent from one edit to the next. OpenAI says both are about 50% faster than 2.5.

Two new output formats come with it:

  • SVG: editable vector files, useful for icons, badges and stickers.
  • Animated GIF: short loops from a text prompt. If a GIF's frame size can't be put together, Sumi Studio tells you before you submit, and you aren't charged.

GPT Image edits and inpainting now keep the aspect ratio of the source image by default, so the frame doesn't change when you edit.

An animated GIF made with GPT Image 2.6 from a text prompt: a woman walking down a colorful cartoon street.

Edit by markup in Draw

Sometimes it's easier to show what to change than to describe where it is. In Draw, mark the area on the image with the brush or the box, write what should change, and GPT Image 2.6 makes the edit. In the example below, we circled Sumi's drink, wrote "change to blue drink", and the rest of the picture stayed the same.

Edit by markup: the drink is circled on the left with the prompt "change to blue drink", and on the right GPT Image 2.6 has turned it blue while the rest of the picture is unchanged.

Seedream 5.5

ByteDance's newest image model, Seedream 5.5, is now available for both text to image and image editing.

Seedance 2.5 Flash

Seedance 2.5 Flash is a faster, cheaper version of Seedance 2.5. We gave both the same 20-second prompt at 720p. Flash finished 31% faster and cost 35% less. Seedance 2.5 kept the man's face the same in every shot and followed the prompt to the end, while Flash changed his face between shots and got some details wrong.

Use Flash for drafts and for testing ideas quickly, and Seedance 2.5 for the final version.

Seedance 2.5 in 4K

Seedance 2.5 can now render at 4K, for when the final cut needs full detail.

Qwen 4 Max

Qwen 4 Max is Alibaba's newest flagship language model. Alibaba built it for coding, knowledge work and long multi-step agent tasks, and says it improves on Qwen 3.8 in world knowledge, code generation and staying consistent over long tasks. You can pick it for the Sumi agent and in Text nodes.

A refreshed workspace

The home page now splits projects and canvases into two panels. You can drag them to change their position and width. Below them, new Explore tabs show recent posts, Sumi Academy and Community.

The new home page: the prompt box at the top, and the What's new, Use cases, Academy and Community tabs below it.

Sumi now sits as a card on the right side of the screen. On wide screens, the page makes room for it so it doesn't cover your work. You can also choose how Sumi's thinking process is shown in conversations.

The product page has a new look too.

The redesigned product page.

Better MCP connections

Agents connected over MCP can now see prices when they look up models, and can read your asset files directly.

Try the new models on Sumi Studio