Mixedbread

Embedding Models

Introduction

is our flagship embedding model family. Enjoy easy access and stellar performance that can help you elevate your retrieval pipeline. Use embeddings for search, classification, recommendation, and other impactful tasks.

What's new in the Mixedbread embed family?

The mixedbread embed family has recently seen exciting developments:

Model Family

Here's an overview of our current model lineup:

ModelCardContext LengthDimensionMTEB Average
512102464.68
5121024 (base)63.25 (base)
5121024-
409638442.80

Why Choose Mixedbread Embeddings

The Mixedbread embed family offers several advantages:

  1. Powerful Performance: State-of-the-art results on benchmarks
  2. Size Efficiency: Optimized for resource utilization
  3. Open-Source: Fully accessible and customizable
  4. Versatility: Suitable for various NLP tasks

Performance Comparison

Our new model outperforms other similarly sized open models and even surpasses some closed-source models on the MTEB benchmark:

ModelAvg (56 datasets)
mxbai-embed-large-v164.68
bge-large-en-v1.564.23
jina-embeddings-v2-base-en60.38
OpenAI text-embedding-3-large (Proprietary)64.58
Cohere embed-english-v3.0 (Proprietary)64.47

API Benefits

While you can use our open-source models directly, our API offers additional advantages:

  1. Enhanced Performance: API-exclusive versions offer improvements like better int8-quantization and take advantage of our optimized inference pipeline.
  2. Calibration Data: Generated using over 50 million samples for more accurate float32 to int8 mapping
  3. Faster Response Times: Optimized for low-latency retrieval tasks
Use Cases
Discover specific products and features you can build with Embeddings.
mxbai-embed-large-v1
Discover mxbai-embed-large-v1, our state-of-the-art English embedding model. Learn about its powerful performance, versatility across various NLP tasks, and how to effectively use it for semantic search, information retrieval, and other applications.

Last updated: June 10, 2025