Hey ​@Riddhima Sharma - happy to get some clarification over to you!

First -- when you mention

Also, what’s the memory capacity of Figma Make? I noticed it was using Calude Sonnet so, assumed it should have been good. Am I missing something here?

Are you referring to the input token limit, or memory in the sense of the recollection of previous instructions?

Also, for the performance experience: our engineering team is really interested in being able to try to replicate what you’re experiencing when it comes to the slow/lagging with the actions you’re taking.

Would you be willing to share any of the files you’re working on that you see this happening on? If so, please reply and let me know here, and I’ll get you connected to our technical quality team for further instructions.

Thank you!

I’m seeing basically the same thing. I’ve been using Figma Make to do some large site mockups, however the deeper I’ve gotten into it, the slower it has gotten. It is to the point where I cannot type in the box directly anymore and am writing prompts in another text editor. I think the more revisions it creates, the worse it gets. I’m guessing the context window is too big. I’m currently at about 70 revisions.

I’d be happy to work with someone to help make this better.

I can attest to the same. I’ve been working on a large-scale mock-up for some time now and I’m currently 873 revisions in. It’s been painful for 700 of those revisions. It’s been unusable for 200 revisions and I’ve been stubbornly pushing forward. The project has been so close to completion, but it’s so unusable now that I can’t finish the project.

What would be amazing is a way to clean up/clear the prompt history. I think that’s what’s causing memory issues because I don’t think the files it’s created are that large and I’ve also prompted multiple times to clean up the code base so try and lower the overall file sizes. When other users try to open the Make file, they struggle to access it because their browser will crash. Sometimes they’re successful after multiple attempts.

In both the VS Code Claude Code extention, and also in Figma Make, I’ve found it helpful to have it write a file, CLAUDE.md in the root of the project directory and tell it to make updates to the file periodically as a way to maintain memory and context. After you complete some work, prompt it like, “Update CLAUDE.md documenting our most recent work.” or something similar.

That said, I would like to know what the token limits are and to understand if I hit that limit today? I’ve been building a prototype of our app this week and have had a lot of success with it. The chatbox all week has reported, “Sonnet 4”. I quit Figma earlier due to rebooting my machine and when I came back to it, now the chat box does not report a model at all. The prompts that I tried produced fairly awful results and it continually failed to read, ‘App.tsx’ and said that the code in the file was the default “Hello, World!” file. That clearly was not true when I verified the content in the actual file in the Code view.

I have been dropping large requests into Figma make for a project. The context is large due to specificity of design/layout, but frankly nothing too crazy. I do have Supabase integrated, and I am seeing an 11 warnings error regarding some of the Supabase lines of code. I try to fix those - it gets hung and never progresses past “Thinking” and the project itself is hidden behind the “Smoothing things out” view.

This started last night and is persistent as of this morning. The issue persists regardless of if I us Figma’s web ui or native app, and it doesn’t matter what request I send it. I have closed and reopened windows to resolve as well.

Looking for support as this is putting me behind my current timeline to meet my current deadline. Please help ASAP.

 Update: I have had 1 request complete after a significant amount of time ~2 hours

Update 2: Copying the project does not resolve this issue.

Hey everyone - I know this is a slightly older topic, but I do want to make sure each of your experiences with the prompt generation (and performance with it) is reviewed individually.

There could be a common denominator across each of your works that we can surface.

If you can: could you each please share the URL of the file where you’re seeing issues (like if there’s a particular one that runs prompts extra, extra slowly), and grant edit access to them with support-share@figma.com by filling out this quick form: https://figma.bot/submitreport. You won’t be charged for this, and will help us reproduce and investigate this more effectively. 

Let me know when you’ve done that (either here or in a PM) and I can get the conversation with support started.