AIHacker Newsabout 1 hour ago
SketchAgentBench Evaluates LLMs on 3D Coding
2 min read
Researchers launch SketchAgentBench, a benchmark testing large language models on 3D spatial reasoning and coding within a browser three.js playground. The evaluation compares 11 models across 560 pairs of photos, anchored at Claude Sonnet 5.5 with a score of 1000.
Level
Hype check
Tap to vote and see what everyone thinks.