Three labs, one council
Claude Fable 5.1, Codex GPT-6 Astra, and Grok 4.6 take the same seat at every run. Nothing to configure.
Sealed multi-model council v1.5
Three frontier models, one from each lab. They work on the same question independently—without seeing one another—then Claude fuses the strongest thinking into one clear result.
Free · Claude Code plugin · Your own model subscriptions
How it works
Claude Fable 5.1, Codex GPT-6 Astra, and Grok 4.6 take the same seat at every run. Nothing to configure.
Every model receives the same brief in a fresh session. No one can copy or anchor on another answer.
The answers are anonymised before judging, so the strongest reasoning wins—not the biggest model name.
Two ways to use it
Use Fusion when one AI answer is too easy to trust. It is built for decisions where blind spots matter and projects where several strong plans should compete before you commit.
Install
Paste the three lines into Claude Code. They add the marketplace, install the plugin, and reload plugins once.
/plugin marketplace add gabrieljudah/llm-fusion /plugin install llm-fusion@fusion /reload-plugins
Already installed? Run /plugin marketplace update fusion, then /plugin update llm-fusion@fusion, then /reload-plugins. Refreshing the marketplace alone does not replace the installed plugin.
Install LLM Fusion and let three labs disagree before you act.