Skip to content
AI.info

companies

Mixedbread

Berlin-based AI lab building open-source embedding and reranking models plus a hosted search API for retrieval and RAG applications.

Mixedbread

Mixedbread was founded in 2023 in Berlin by Aamir Shakir and Julius Lipp, who set out to build open-source embedding and reranking models competitive with closed offerings from larger AI labs. The company's mxbai-embed-large-v1 model, released under an Apache 2.0 license, scored ahead of OpenAI's text-embedding-ada-002 on the MTEB benchmark at launch and has since passed several million downloads on Hugging Face, making it one of the more widely adopted open embedding models in its size class. Beyond the open-weight models, Mixedbread runs a hosted API business covering embeddings, reranking, and a broader multimodal search and vector store product, aimed at developers who want managed infrastructure without giving up the option to self-host the underlying open-source models. The reranker lineup spans from open mxbai-rerank models to a proprietary listwise reranker and compact edge-optimized variants for latency-sensitive deployments. The company has stayed small and largely bootstrapped relative to peers in the retrieval space, raising a modest seed round in 2024 while building its reputation primarily through open-source releases and community adoption rather than large funding announcements.

Founded
2023
Headquarters
Berlin, Germany
Sector
infrastructure

Tools

  • Mixedbread Rerank

    Reranks an initial set of search results with specialized models to improve relevance for complex queries.

  • Mixedbread Embed

    Legacy direct embedding APIs and models from Mixedbread; new integrations are directed to Mixedbread Search Stores.

Official website