Glossary
The terms behind multi-model AI
Plain-English definitions of the concepts that come up when you run more than one model. No jargon for its own sake, just what each term means and why it matters.
Best-of-N
Generating several candidate answers and keeping the best one, instead of trusting a single attempt.
Consensus
The points that several models agree on, which tend to be more reliable than any single model’s claims.
Context window
The amount of text a model can consider at once, measured in tokens, covering both your input and its output.
LLM as a judge
Using one model to evaluate, rank, or merge the outputs of other models.
LLM orchestration
Coordinating one or more large language models, and the steps around them, to complete a task that a single prompt would not handle well on its own.
Mixture of experts
A model architecture that activates only a relevant subset of its parameters for each input, giving large-model capability at lower running cost.
Model ensemble
Combining the outputs of several models into one result, so the group performs better than any single member.
Model routing
Sending each task, or each part of a task, to the model best suited for it, rather than using one model for everything.
Multi-agent debate
Having models argue different positions on a question, then settle on a conclusion, so the reasoning is tested rather than taken at face value.
Multi-model AI
Using more than one AI model on the same task, instead of relying on a single model for every answer.
Open-weights model
A model whose trained weights are released publicly, so it can be run and hosted by anyone, not just its original lab.
Reasoning model
A model built to work through problems step by step before answering, rather than responding in a single pass.
Response synthesis
Merging several model answers into one clean result, rather than showing them side by side and making you pick.
Tokens
The chunks of text models read and write, roughly a few characters each, and the unit most model pricing is based on.