Model Council is the best way to get one question answered by several labs and summarised. Bordet is for the project that does not end with one answer.
Perplexity's Model Council does one thing very well. You pick two to eight models, they answer your question independently, and a chair model you choose writes a synthesis: where they agree, where they disagree, what each saw that the others missed. It is included in Pro at $20 a month and in Computer it can produce reports and decks. For a single high-stakes question, it is excellent, and we say so.
A council convenes for a question. The models never read each other; they answer alone and the chair reconciles. When you come back tomorrow with the follow-up, you convene a new council, and nothing the last one concluded binds this one. There is no board, no decisions log, no room another person or your own AI can walk into.
Bordet is a standing table. The same models, GPT-6 Astra, Claude Fable 5.1, Gemini 3.1 Pro, Grok 4.6, sit in one thread with roles and read each other's replies, so a disagreement is a real exchange, not two answers laid side by side. The room keeps one memory that every model and every person reads: objective, status, open questions, decisions you confirm, sources cited by name, deliverables with versions. Your teammates and your own claude.ai, ChatGPT or Claude Code can take a seat. The argument continues over weeks, and the table holds itself to what it decided.
| Capability | Bordet | Perplexity Model Council |
|---|---|---|
| Models from several labs | GPT, Claude, Gemini, Grok, DeepSeek, one model per member, any mix | 2 to 8 per query from OpenAI, Anthropic, Google and open-weight labs |
| Do the models read each other? | Yes, one shared thread; they challenge and build on each other by name | No, independent answers reconciled by a chair |
| A written synthesis of agreement and disagreement | Not yet: the argument is on the page, and you record the outcome on the board | Yes, from the chair model, per query |
| Continuity between questions | One project memory across sessions, read by every model | Each council is a new query; Spaces hold files and threads |
| Other people in the same conversation | Yes, live, with invite links and guests | Not in a council |
| Your own AI tools at the table | claude.ai, ChatGPT, Claude Code and Cursor connect as seats | n/a |
| A record of decisions | Decisions log with confirm, plus a board of objective, status and open questions | No |
| Sources | Links, PDF and Word files, notes, read and cited by every member; live web search per member | Web search and Spaces files |
| Pricing | Free $2 to start, credits from $5, or bring your own keys; per-room spend caps | Included in Pro ($20/mo) and Max ($200/mo) |
Comparison reflects our understanding as of September 2026; check Perplexity Model Council's site for their current features.
Bordet has no subscription. Start with $2 free, then buy credits from $5 and spend them only when a member speaks, or bring your own API keys and pay the labs directly with no markup. A focused session at today's frontier models costs about a dollar or two managed. Per-room spend caps keep it predictable.
Because today the exchange itself is the product: you see Claude push back on GPT's number and GPT concede or defend it. You record the outcome on the board and in the decisions log, and the table holds itself to that next time. A written round-up is on our list; when it ships it will sit on top of the argument, not replace it.
Yes. GPT-6 Astra, GPT-5.5, Claude Fable 5.1, Opus 5, Gemini 3.1 Pro, Gemini 3.8 Flash, Grok 4.6 and DeepSeek V4 Pro are in the catalog, and you choose which one holds which role.
Many people do: convene a council for a one-off question, and bring the answer into the Bordet room as a source so the table can work with it and remember it.
Open a boardroom, give each model a role, and watch the team remember what you decided.