Reasoning Council — Claude Opus 4.7, GPT-5.1, Gemini 3 Pro, Grok 4. Deep, multi-step reasoning.
Code Review — Claude Sonnet 4.5, GPT-5.1, DeepSeek V4, Qwen 3.6 Plus. Catches bugs across paradigms.
Research — All frontier models. Maximum breadth for literature review and fact-checking.
Cost-Aware — Kimi K2.5, Qwen 3.6 Plus, DeepSeek V4, GLM 5.1. Solid quality at minimal cost.
Frequently asked questions
Which models are in the default council?
GPT-5.1, Claude Sonnet 4.5, Gemini 3 Pro, Grok 4, Kimi K2.5, Qwen 3.6 Plus, and GLM 5.1. You can swap any of them, add others from the catalog, or save your own preset.
Can I bring my own API keys?
No — Counsel runs through a single billing relationship to keep usage predictable. Per-model token costs are passed through transparently in the dashboard.
How many models can be in one council?
Up to 12 council members plus 1 chairman — the same limit applies across every plan. The chairman synthesises the council's responses and is never also a member.
Do you support open-source models?
Yes — Llama 3.1 405B, Mistral, DeepSeek V4, Qwen, GLM, and Kimi are all available. Mixing closed and open in the same council is a common pattern for cost-aware quality.