Sol Outperforms Fable in Drawing Code Generation Benchmark

suchenzang · x · 2026-08-23

User compared the performance of Sol and Fable models on generating drawing code (DT code) and counting crossings. Sol performed significantly better in this task. While Fable managed to capture pen stroke details (an "A+ effort"), it tended to get stuck on irrelevant details and failed to accurately count crossings or produce correct code, even when provided with manually improved, clean SVG inputs.

Original post →

More from Models

Models channel →