Figma details training and evaluation for its dedicated canvas editing model
What happened
On October 30, Figma described how it trains and evaluates a dedicated canvas editing model, focused on making precise edits and iterations to existing designs through supervised fine-tuning and reinforcement learning. Training uses native file formats, synthetic edit requests and successful examples. Evaluation combines structural checks, vision-model judging and designer calibration. Figma's newest editing model is not yet released; it is post-trained on Qwen 3.8 27B.
Written by AI from the coverage · updated 2 hours ago
Coverage
Follow the reports to see the story from different sides.
- AI HOT · Tips & opinionsFigma 讲解如何训练专用画布编辑模型
Figma 讲解了专用画布编辑模型的训练与评估方法，通过监督微调和 RL 改善既有设计的精准修改与迭代。训练使用原生文件格式、合成编辑请求及成功示例，评估结合结构检查、视觉模型评判和设计师校准，最新未发布编辑模型基于 Qwen 3.8 27B 后训练。
Heat over time
Not enough continuous observations to draw a trend yet.