We have upgraded the core model behind process-map generation.

Maps now arrive around 2.8x faster on our standard benchmark: roughly 3 seconds for a typical process map, down from about 9.

The bigger win is on long briefs. Paste in a detailed procedure and you are far more likely to get a complete set of three options back rather than a partial one.

We did not trade quality for speed. Before shipping it we ran a blind evaluation across the same task set, scored by independent judges that could not see which model produced which map. The new model matched the old one.

Faster, same quality, nothing to switch on.