AI Potluck
Contribute
Back to Gap Map Model components / Embeddings & retrieval models

mxbai-embed

Mixedbread
open weights / Overall score: 3.3

mxbai-embed is Mixedbread's open-weight English sentence embedding line, headed by mxbai-embed-large-v1, a BERT-large bi-encoder with a 512-token window, released in March 2024. The family also includes a 2D-Matryoshka variant, an xsmall checkpoint and a German model built with deepset. All of them support Matryoshka truncation and binary quantization, which is the line's distinguishing feature: a 1024-dimension float vector can be cut to a fraction of its storage cost with most of its retrieval quality intact.

Openness

3 high confidence
3.0
weights
open(four ungated Apache-2.0 checkpoints)
data
closed(training data not published
code
partial(inference, Matryoshka truncation and binary-quantization snippets on the card)
license
Apache-2.0(OSI)

Every checkpoint ships under Apache-2.0, ungated. The training corpus is the gap: the card's only claim about the data is that it does not overlap MTEB, which is about evaluation hygiene rather than disclosure, and neither the corpus nor training code is published.

Adoption

4 high confidence
4.0

Adoption is measured as Hugging Face downloads across the four embed checkpoints, almost all of it for mxbai-embed-large-v1 with smaller shares for the 2D-Matryoshka, deepset German and xsmall variants. Mixedbread's rerank and colbert lines are excluded as separate products.

Capability

3 high confidence
3.0

mxbai-embed-large-v1 is an earlier-generation model — English-only, a 512-token window — with an MTEB v1 average that has never been re-run on v2, so it cannot be measured against newer embedding models on the same benchmark. It is scored level with Granite Embedding, the nearest comparable peer: a similarly compact bi-encoder, one generation newer, with sixteen times the context.

Verified 2026-09-11