Configuration saved and applied to chat.

What is a Pipeline?

Instead of asking all models the same question at once, a pipeline sends your request through models one at a time, in order. Each model sees what the previous one produced and builds on it. Like an assembly line — research first, then strategy, then execution.

When ON, models respond one at a time in order. Each model sees what the previous one said. Like a real conversation — slower but more thoughtful. When OFF, all models respond at the same time independently.

When ON, every model can see what ALL other models have said. They build on each other's answers. When OFF, each model only sees YOUR messages and its own previous replies — they work in isolation.

When ON, every question you ask gets searched on the web first (Exa Deep Max). The top 5 sources are fed to every model as context before they answer. Models will cite source numbers [1][2] in their replies. Adds about 1 second per turn and a few cents of Exa usage.

Shows how long each model took to respond (e.g. "2.3s"). Useful for comparing speed between models.

Shows how many tokens were used per response (e.g. "142 in / 387 out"). Tokens = the units AI models use to process text. More tokens = more cost.