My experience with the agent has been slow and not very good - at least with basic asks over the last few weeks. I like Claude’s performance much better but of course not as simple to integrate the two but certainly possible.
….can you share what is the foundation of Figma’s agent and how is it improving?
Hi
I completely understand the frustration with the AI Agent in Figma Design. Here's what I know: when Figma Make first launched in May 2025, it was built on Claude 3.7 Sonnet. Since then, the team has introduced additional models. More info here. As for the broader Figma Agent LLM foundation and whether the model stack has evolved further, I'm checking with the team to make sure I give you the most accurate and up-to-date answer. I'll follow up here when/if I hear back, but it is actually likely to also be Claude.
In the meantime, a couple of questions that may help us understand your experience better:
Agent performance is something the team is actively working on, so your input genuinely helps. Thanks again for taking the time to share it!
Enter your E-mail address. We'll send you an e-mail with instructions to reset your password.