Hi @artalar Just wanted to follow up here: The “Multiple models” picker was actually removed from the IDE / Agents Window earlier this year. It briefly reappeared in 3.0.9, but that was a regression from the worktrees rollout. It wasn’t actually dispatching parallel runs end-to-end, so we pulled it again.
You’re right that /best-of-n is solving a different problem too — it picks a winner for you rather than letting you compare runs side by side.
The closest workflow today is launching the same prompt across different models as cloud agents from cursor.com/agents, or as separate local agents in the Agents Window with 3.2’s tiled layout. It’s manual (one launch per model), but each run stays independent, and you can inspect each one on its own.