Hasty Briefsbeta

Bilingual

Qwen3.6-35B-A3B on my laptop drew me a better pelican than Claude Opus 4.7

4 hours ago
  • #Benchmark
  • #SVG Generation
  • #AI Models
  • Qwen3.6-35B-A3B generated a better SVG of a pelican riding a bicycle than Claude Opus 4.7, which had issues with the bicycle frame.
  • The author also tested both models on 'a flamingo riding a unicycle' and again found Qwen's output superior.
  • The pelican benchmark was originally intended as a joke about the difficulty of comparing AI models, but quality often correlated with general usefulness.
  • This time, the correlation broke: Qwen's smaller quantized model outperformed Anthropic's latest proprietary model on this specific task.