Ask several models at once and pick the best answer

One task, several AI models from different companies, and a lead session that compares their answers.

Useful when the decision is not obvious: architecture, finding the cause of a bug, choosing a library. Different models make different mistakes, and comparing them shows the weak spots.

Step by step

Open the session that will lead.
Click ⚡ and choose “Model council”.
Turn on the members. The list only shows what is installed on this Mac: ChatGPT · Codex, Gemini, OpenCode and “A second Claude”.
Click “Put into the field”, add your task and send it.
The lead first solves the task on its own, then hands it to the members. Each member appears as a row under the lead in the sidebar.
When everyone has answered, the lead writes a short table: who proposed what, strong and weak points. Then its own choice and why.

Good to know

A council spends the limits of every model involved. If the task is large, the lead warns you first and suggests narrowing it.
Members only read and reply. Only the lead edits project files, after the verdict.
“A second Claude” has not seen your conversation, so it does not lean toward the lead.
Members get the task with all the context they need: they do not see your conversation.
Factual disputes are settled by checking the code, not by counting votes.
Claude and Codex members run as incognito: their conversations are erased after the council.
When the lead finishes its turn, the members close on their own. A new council starts them again.
You can fold the member rows with the arrow on the lead’s row.

If something does not work

A member you want is not in the list: install it in “Agents and models”. See agents.
A member takes a long time: Hedoq waits up to 20 minutes for each member’s answer.