Best open source LLM for performing image analysis of design files?
I’m a product designer who’s playing around with various LLMs to see how they could potentially fit into my workflow. Currently, I’ve been playing around with having GPT Images generate images detailing UI component design specs, and then asking Codex to read the specs and implement them. However, this runs through my limits pretty quickly, so I’m looking to see if any of the open source LLMs could potentially work here.
I originally looked at using Deepseek, but it can’t read images. Design Arena has Kimi and GLM trading blows, so I was wondering if anybody has experience with using them for implementing UI components either from an image, or just in general. Also looked at Qwen but it doesn’t show up in Design Arenas benchmarks too often. Any advice would be appreciated!