Doubao Pro tops a 10-question model test

Doubao Pro topped the September 26 YZ Index Smoke test at 92.85, edging GPT-5.5 at 91.32 and Claude Opus 4.7 at 89.47 across 10 models. The run’s benchmark brief says Smoke weights code execution 55% and material constraints 45%, with only 10 questions; its daily score is a monitoring signal, not a replacement for weekly rolling rankings. Teams comparing reported benchmark splits should record question set and weighting before treating Doubao’s lead as a model-selection signal.