Matt Wolfe created BuseyBench, a benchmark that asks AI models to generate SVG code drawing actor Gary Busey's face. GPT-4 and GPT-3.5-Turbo produced quirky results, and the tool lets you compare newer models' outputs sorted by cost, tokens, and speed.
I built a benchmark where AI models have one job: draw Gary Busey using code. It's called BuseyBench, and it's exactly as ridiculous as it sounds. So the test is simple. I asked different AI models to draw Gary Busey as an SVG. Now, it's not a normal AI image. It's not using Nanobanana or DALL-E or Stable Diffusion or anything like that. SVGs are actually code. The model has to write shapes and lines that somehow become Gary Busey, and that makes the results kind of awesome. Back in March 2023, GPT-4 GPT-3.5 Turbo had its own very special interpretation of Gary Busey. Then you scroll forward and you can watch the models evolve at making SVGs. Some get better, some get weirder, and well, some just kind of chose chaos. I also built in sorting so you can compare cost and tokens used and how long each run took. And there's a timeline page so you can filter providers and watch their Gary Busey journey. The whole benchmark and website was built using Fable, which feels kind of appropriate because this is really, really dumb, but it's actually kind of useful.
AI News: Dots, GPT-6.1 Sol, Sonnet 5.5, Gemini 4, and everything you need to know
