mteb-add-model-10-lines-code

IN premise — summaries/2026/08/24/muennighoff-2022-mteb-s1-introduction.md

Created 2026-08-25T02:58:18+00:00

Adding a model to the MTEB benchmark requires fewer than 10 lines of code; the model must accept a list of texts and return a fixed-shape vector per item.

Summary

Evaluating a new text-embedding model against this benchmark is almost trivial: you just need a function that takes a list of texts and spits out a fixed-size vector for each one, and the whole integration is under ten lines of code. That keeps the barrier to entry very low, so a wide range of models can be compared on equal footing without heavy engineering overhead.