selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

Best open-weight Rerankers models

Cross-encoders that re-score search results to improve RAG and search quality.

Rerankers are cross-encoder models that re-score search results to improve retrieval-augmented generation (RAG) and search quality. They take a set of candidate results and reorder them by relevance to a query. This list ranks open-weight reranker models you can download and run on your own hardware by popularity. Among the popular options are Qwen3-Reranker-0.6B from Qwen, bge-reranker-v2-m3 from BAAI, and jina-reranker-v3 from jinaai. Qwen3-Reranker-4B and Qwen3-Reranker-8B are larger models from the same Qwen family.

--sort
33 models
BAAI logo≈1.8 GB

Different from embedding model, reranker uses question and document as input and directly output similarity instead of embedding. You can get a relevance score by inputting query and passage to the reranker. And the score can be mapped to a float value in [0,1] by sigmoid function.

568M8K ctx Apache 2.0
14.6MMar 2024gguf
Qwen logo≈1.8 GB

The Qwen3 Embedding model series is the latest proprietary model of the Qwen family, specifically designed for text embedding and ranking tasks.

596M40K ctx Apache 2.0
1.2MMay 2025gguf
Qwen logo≈5.3 GB

The Qwen3 Embedding model series is the latest proprietary model of the Qwen family, specifically designed for text embedding and ranking tasks.

4B40K ctx Apache 2.0
2.6MJun 2025gguf
BAAI logo≈1.1 GB

We have updated the new reranker, supporting larger lengths, more languages, and achieving better performance.

278M514 ctx MIT
2.3MSept 2023gguf

The Qwen3-VL-Embedding and Qwen3-VL-Reranker model series are the latest additions to the Qwen family, built upon the recently open-sourced and powerful Qwen3-VL foundation model.

2.1B256K ctx Apache 2.0
85.4KJan 2026gguf
Qwen logo≈10 GB

The Qwen3 Embedding model series is the latest proprietary model of the Qwen family, specifically designed for text embedding and ranking tasks.

8.2B40K ctx Apache 2.0
78.2KMay 2025gguf
jinaai logo≈1.8 GB

jina-reranker-v3 is a 0.6B parameter multilingual document reranker with a novel last but not late interaction architecture.

597M128K ctx CC BY-NC 4.0
745.4KSept 2025gguf

The Qwen3-VL-Embedding and Qwen3-VL-Reranker model series are the latest additions to the Qwen family, built upon the recently open-sourced and powerful Qwen3-VL foundation model.

8.8B256K ctx Apache 2.0
11.6KJan 2026gguf
Alibaba-NLP logo≈0.8 GB

We are excited to introduce the gte-modernbert series of models, which are built upon the latest modernBERT pre-trained encoder-only foundation models. The gte-modernbert series models include both text embedding models and rerank models.

150M8K ctx Apache 2.0
1.4MJan 2025gguf
jinaai logo≈1.8 GB

jina-reranker-v3.5 is a 0.6B-parameter multilingual listwise document reranker and a drop-in upgrade to jina-reranker-v3. It keeps the same last-but-not-late (LBNL) listwise interface — ranking a query against many candidates in one forward pass — while improving domain robustness, structured-data…

597M128K ctx CC BY-NC 4.0
169.5KJul 2026gguf
BAAI logo≈1.8 GB

We have updated the new reranker, supporting larger lengths, more languages, and achieving better performance.

560M514 ctx MIT
2MSept 2023gguf

The Jina Reranker v2 (jina-reranker-v2-base-multilingual) is a transformer-based model that has been fine-tuned for text reranking task, which is a crucial component in many information retrieval systems.

278M1K ctx CC BY-NC 4.0
1.5MJun 2024gguf

The Llama Nemotron Reranking 1B model is optimized for providing a logit score that represents how relevant a document(s) is to a given query. The model was fine-tuned for multilingual, cross-lingual text question-answering retrieval, with support for long documents (up to 8192 tokens).

1.2B128K ctx openmdw-1.1
297.4KOct 2025gguf
Alibaba-NLP logo≈1.2 GB

The gte-multilingual-reranker-base model is the first reranker model in the GTE family of models, featuring several key attributes: - High Performance: Achieves state-of-the-art (SOTA) results in multilingual retrieval tasks and multi-task representation model evaluations when compared to reranker…

306M8K ctx Apache 2.0
41.7KJul 2024
mixedbread-ai logo≈4.0 GB

🍞 Looking for a simple end-to-end retrieval solution? Meet Omni, our multimodal and multilingual model. Get in touch for access.

1.5B32K ctx Apache 2.0
10.4KMar 2025gguf
jinaai logo≈6.0 GB

jina-reranker-m0 is our new multilingual multimodal reranker model for ranking visual documents across multiple languages: it accepts a query alongside a collection of visually rich document images, including pages with text, figures, tables, infographics, and various layouts across multiple…

2.4B32K ctx CC BY-NC 4.0
71.9KMar 2025gguf

Nemotron-3.5-Rerank-8B-BF16 was developed by NVIDIA for multilingual and multimodal question-answering retrieval. It is a world-class reranker optimized to produce a logit representing the relevance of a document to a text query.

8.4B256K ctx openmdw-1.1
8.2KOct 2026gguf
Contrastive-LM logo

CLM-v0.1-8B

Contrastive-LM

≈10 GB

Contrastive Language Model (CLM) is a new class of System One model trained with a contrastive learning objective that connects states and actions.

8BApache 2.0
1.1KSept 2026gguf
mixedbread-ai logo≈1.6 GB

🍞 Looking for a simple end-to-end retrieval solution? Meet Omni, our multimodal and multilingual model. Get in touch for access.

494M32K ctx Apache 2.0
16.5KMar 2025gguf

Different from embedding model, reranker uses question and document as input and directly output similarity instead of embedding. You can get a relevance score by inputting query and passage to the reranker. And the score can be mapped to a float value in [0,1] by sigmoid function.

2.5B8K ctx Apache 2.0
105KMar 2024gguf
mixedbread-ai logo≈0.7 GB

🍞 Looking for a simple end-to-end retrieval solution? Meet Omni, our multimodal and multilingual model. Get in touch for access.

71M512 ctx Apache 2.0
608.7KFeb 2024gguf

This model is designed for blazing-fast reranking while maintaining competitive performance. What's more, it leverages the power of our JinaBERT model as its foundation. JinaBERT itself is a unique variant of the BERT architecture that supports the symmetric bidirectional variant of ALiBi.

33M8K ctx Apache 2.0
17.6KApr 2024gguf
zeroentropy logo≈5.3 GB

In search engines, rerankers are crucial for improving the accuracy of your retrieval system.

4B40K ctx Apache 2.0
68.2KNov 2025gguf

Model Summary: granite-embedding-reranker-english-r2 is a 149M parameter dense cross-encoder model from the Granite Embeddings collection that can be used to generate high quality text embeddings. This model produces embedding vectors of size 768 based on context length of upto 8192 tokens.

150M8K ctx Apache 2.0
12.1KAug 2025gguf