• open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
• embedding — A list of numbers representing a piece of text's meaning, so that similar meanings end up numerically close and can be searched.
• quantization — Shrinking a model by storing its numbers less precisely — like rounding — so it runs faster and fits on smaller hardware, at a small quality cost.

You can run embeddings locally during development and move to the hosted Voyage 4 models later without re-indexing, because the vectors sit in the same space.

• postVoyage's rerankers start taking instructions at query time
• postWhy code retrieval benchmarks mislead, and how to rebuild them
• postWhen embedding a chunk should include the whole document
• postLong context windows do not fix LLM reranking

• postVoyage 4 shares one embedding space across large, lite and nanoVoyage AI by MongoDB
• postVoyage 4 puts every model size in one embedding spaceVoyage AI by MongoDB
• postThree levers for cutting retrieval cost in productionVoyage AI by MongoDB

Checking sign-in…

Loading comments…