mdbr-leaf-mt is a compact high-performance text embedding model designed for classification, clustering, semantic sentence similarity and summarization tasks.

To enable even greater efficiency, mdbr-leaf-mt supports flexible asymmetric architectures and is robust to vector quantization and MRL truncation.

If you are looking to perform semantic search / information retrieval (e.g. for RAGs), please check out our mdbr-leaf-ir model, which is specifically trained for these tasks.

[!Note]
Note: this model has been developed by the ML team of MongoDB Research. At the time of writing it is not used in any of MongoDB's commercial product or service offerings.

Technical Report

A technical report detailing our proposed LEAF training procedure is available here.

Highlights

State-of-the-Art Performance: mdbr-leaf-mt achieves new state-of-the-art results for compact embedding models, ranking #1 on the public MTEB v2 (Eng) benchmark leaderboard for models with ≤30M parameters.
Flexible Architecture Support: mdbr-leaf-mt supports asymmetric retrieval architectures enabling even greater retrieval results. See below for more information.
MRL and Quantization Support: embedding vectors generated by mdbr-leaf-mt compress well when truncated (MRL) and can be stored using more efficient types like int8 and binary. See below for more information.

Benchmark Comparison

The table below shows the scores for mdbr-leaf-mt on the MTEB v2 (English) benchmark, compared to other retrieval models.

mdbr-leaf-mt ranks #1 on this benchmark for models with <30M parameters.

Model	Size	MTEB v2 (Eng)
OpenAI text-embedding-3-large	Unknown	66.43
OpenAI text-embedding-3-small	Unknown	64.56
mdbr-leaf-mt	23M	63.97
gte-small	33M	63.22
snowflake-arctic-embed-s	32M	61.59
e5-small-v2	33M	61.32
granite-embedding-small-english-r2	47M	61.07
all-MiniLM-L6-v2	22M	59.03

Quickstart

Sentence Transformers

python

from sentence_transformers import SentenceTransformer  
  
# Load the model  
model = SentenceTransformer("MongoDB/mdbr-leaf-mt")  
  
# Example queries and documents  
queries = [
    "What is machine learning?",  
    "How does neural network training work?"  
]  
  
documents = [  
    "Machine learning is a subset of artificial intelligence that focuses on algorithms that can learn from data.",  
    "Neural networks are trained through backpropagation, adjusting weights to minimize prediction errors."  
]  
  
# Encode queries and documents  
query_embeddings = model.encode(queries, prompt_name="query")  
document_embeddings = model.encode(documents)  
  
# Compute similarity scores  
scores = model.similarity(query_embeddings, document_embeddings)  

# Print results
for i, query in enumerate(queries):
    print(f"Query: {query}")
    for j, doc in enumerate(documents):
        print(f" Similarity: {scores[i, j]:.4f} | Document {j}: {doc[:80]}...")

See example output

Query: What is machine learning?
 Similarity: 0.9063 | Document 0: Machine learning is a subset of ...
 Similarity: 0.7287 | Document 1: Neural networks are trained ...

Query: How does neural network training work?
 Similarity: 0.6725 | Document 0: Machine learning is a subset of ...
 Similarity: 0.8287 | Document 1: Neural networks are trained ...

Transformers.js

If you haven't already, you can install the Transformers.js JavaScript library from NPM using:

npm i @huggingface/transformers

You can then use the model to compute embeddings like this:

import { AutoModel, AutoTokenizer, matmul } from "@huggingface/transformers";

// Download from the 🤗 Hub
const model_id = "MongoDB/mdbr-leaf-mt";
const tokenizer = await AutoTokenizer.from_pretrained(model_id);
const model = await AutoModel.from_pretrained(model_id, {
    dtype: "fp32", // Options: "fp32" | "fp16" | "q8" | "q4" | "q4f16"
});

// Prepare queries and documents
const queries = [
    "What is machine learning?",
    "How does neural network training work?",
];
const documents = [  
    "Machine learning is a subset of artificial intelligence that focuses on algorithms that can learn from data.",
    "Neural networks are trained through backpropagation, adjusting weights to minimize prediction errors.",
];
const inputs = await tokenizer([
    ...queries.map((x) => "Represent this sentence for searching relevant passages: " + x),
    ...documents,
], { padding: true });

// Generate embeddings
const { sentence_embedding } = await model(inputs);
const normalized_sentence_embedding = sentence_embedding.normalize();

// Compute similarities
const scores = await matmul(
    normalized_sentence_embedding.slice([0, queries.length]),
    normalized_sentence_embedding.slice([queries.length, null]).transpose(1, 0),
);
const scores_list = scores.tolist();

for (let i = 0; i < queries.length; ++i) {
    console.log(`Query: ${queries[i]}`);
    for (let j = 0; j < documents.length; ++j) {
        console.log(` Similarity: ${scores_list[i][j].toFixed(4)} | Document ${j}: ${documents[j]}`);
    }
    console.log();
}

See example output

Query: What is machine learning?
 Similarity: 0.9063 | Document 0: Machine learning is a subset of artificial intelligence that focuses on algorithms that can learn from data.
 Similarity: 0.7287 | Document 1: Neural networks are trained through backpropagation, adjusting weights to minimize prediction errors.

Query: How does neural network training work?
 Similarity: 0.6725 | Document 0: Machine learning is a subset of artificial intelligence that focuses on algorithms that can learn from data.
 Similarity: 0.8287 | Document 1: Neural networks are trained through backpropagation, adjusting weights to minimize prediction errors.

Transformers Usage

See here.

Asymmetric Retrieval Setup

[!Note] Note: a version of this asymmetric setup, conveniently packaged into a single model, is available here.

mdbr-leaf-mt is aligned to mxbai-embed-large-v1, the model it has been distilled from, making the asymmetric system below possible:

python

# Use mdbr-leaf-mt for query encoding (real-time, low latency)  
query_model = SentenceTransformer("MongoDB/mdbr-leaf-mt")  
query_embeddings = query_model.encode(queries, prompt_name="query")  

# Use a larger model for document encoding (one-time, at index time)  
doc_model = SentenceTransformer("mixedbread-ai/mxbai-embed-large-v1")  
document_embeddings = doc_model.encode(documents)  
  
# Compute similarities  
scores = query_model.similarity(query_embeddings, document_embeddings)  

Retrieval results from asymmetric mode are usually superior to the standard mode above.

MRL Truncation

Embeddings have been trained via MRL and can be truncated for more efficient storage:

python

query_embeds = model.encode(queries, prompt_name="query", truncate_dim=256)
doc_embeds = model.encode(documents, truncate_dim=256)

similarities = model.similarity(query_embeds, doc_embeds)

print('After MRL:')
print(f"* Embeddings dimension: {query_embeds.shape[1]}")
print(f"* Similarities: \n\t{similarities}")

See example output

After MRL:
* Embeddings dimension: 256
* Similarities:
    tensor([[0.9164, 0.7219],
            [0.6682, 0.8393]], device='cuda:0')

Vector Quantization

Vector quantization, for example to int8 or binary, can be performed as follows:

Note: For vector quantization to types other than binary, we suggest performing a calibration to determine the optimal ranges, see here. Good initial values are -1.0 and +1.0.

python

from sentence_transformers.quantization import quantize_embeddings
import torch

query_embeds = model.encode(queries, prompt_name="query")
doc_embeds = model.encode(documents)

# Quantize embeddings to int8 using -1.0 and +1.0
ranges = torch.tensor([[-1.0], [+1.0]]).expand(2, query_embeds.shape[1]).cpu().numpy()
query_embeds = quantize_embeddings(query_embeds, "int8", ranges=ranges)
doc_embeds = quantize_embeddings(doc_embeds, "int8", ranges=ranges)

# Calculate similarities; cast to int64 to avoid under/overflow
similarities = query_embeds.astype(int) @ doc_embeds.astype(int).T

print('After quantization:')
print(f"* Embeddings type: {query_embeds.dtype}")
print(f"* Similarities: \n{similarities}")

See example output

After quantization:
* Embeddings type: int8
* Similarities:
   [[2202032 1422868]
    [1421197 1845580]]

Evaluation

Please see here.

Citation

If you use this model in your work, please cite:

bibtex

@inproceedings{vujanic-ruckstiess-2026-leaf,
    title = "{LEAF}: Knowledge Distillation of Text Embedding Models with Teacher-Aligned Representations",
    author = {Vujanic, Robin  and
      R{\"u}ckstie{\ss}, Thomas},
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2026.acl-long.2008/",
    doi = "10.18653/v1/2026.acl-long.2008",
    pages = "43362--43383",
    ISBN = "979-8-89176-390-6",
    abstract = "We present a knowledge distillation framework for text embedding models. A key distinguishing feature is that our distilled models are compatible with their teacher, enabling flexible asymmetric architectures where documents are encoded with the larger teacher model, while queries use smaller student models. We also show that our models automatically inherit MRL and robustness to output quantization whenever these properties are present in the teacher model, without explicitly training for them. To demonstrate the effectiveness of our framework we publish leaf-ir, a 23M parameters information retrieval oriented model that, besides being teacher-compatibile, sets a new state-of-the-art (SOTA) on BEIR, ranking no.1 on the public leaderboard for models of its size. Asymmetric mode further increases its retrieval performance. Our scheme is however not restricted to information retrieval. We demonstrate its wider applicability by synthesizing the multi-task leaf-mt model. This also sets a new SOTA, achieving no.1 on the public MTEB v2 (English) leaderboard for models of its size. Our technique is applicable to black-box models, requires no judgments nor hard negatives, and training can be conducted using small batch sizes. Thus, dataset and training infrastructure requirements for our framework are modest. We make our models publicly available under a permissive Apache 2.0 license."
}

License

This model is released under Apache 2.0 License.

Contact

For questions or issues, please open an issue or pull request. You can also contact the MongoDB ML Research team at robin.vujanic@mongodb.com.