mxbai-rerank
MixedbreadMixedbread's cross-encoder reranking line, distinct from their mxbai-embed bi-encoders. The v2 generation (base 0.5B, large 1.5B) is trained with a three-step GRPO, contrastive and preference-learning recipe, covers 100-plus languages and 8K-token documents, and reports 57.49 BEIR average for large-v2; the v1 generation (xsmall, base, large) predates it and still carries most of the line's downloads.
Added 2026-09-11 for the embeddings_retrieval promotion. Confirmed as a real and separate line - five mxbai-rerank checkpoints ship under their own PyPI package and GitHub repository, none of which is an mxbai-embed artifact. The mxbai-colbert and mxbai-edge-colbert late-interaction models are a third line again and are not counted here. Verified 2026-09-11 via the mxbai-rerank-large-v2 model card and the mixedbread-ai/mxbai-rerank repository.
Openness
3 high confidence- weights
- open(Apache-2.0 ungated safetensors across the v1 and v2 checkpoints)
- data
- closed(no training data released
- code
- partial(the mxbai-rerank package is inference only)
- license
- Apache-2.0(OSI)
Apache-2.0 ungated weights with an inference package but no training pipeline and no corpus, which is this category open-weights centre of gravity. The three-step GRPO, contrastive and preference-learning recipe is described in prose and in arXiv 2506.03487 rather than shipped.
- https://huggingface.co/mixedbread-ai/mxbai-rerank-large-v2/raw/main/README.md recorded 2026-09-11
Front matter reads license: apache-2.0 with 100-plus language tags; the Training Details section describes a three-step GRPO, contrastive learning and preference learning process and points at arXiv 2506.03487 and the mixedbread blog rather than at released data or training code.
- https://raw.githubusercontent.com/mixedbread-ai/mxbai-rerank/main/README.md recorded 2026-09-11
The mixedbread-ai/mxbai-rerank repository README documents the pip-installable mxbai-rerank inference package and the MxbaiRerankV2 class, describes the models as Apache 2.0 open source, and offers no training entry point.
Adoption
3 high confidenceAbout 804K downloads in the trailing 30 days across the five mxbai-rerank checkpoints - xsmall-v1 631K, large-v2 102K, base-v1 43K, large-v1 31K, base-v2 28K. The mxbai-colbert and mxbai-edge-colbert late-interaction models are a separate line and are not counted.
- https://huggingface.co/api/models/mixedbread-ai/mxbai-rerank-xsmall-v1 recorded 2026-09-11
downloads: 630939; gated: false; private: false for mixedbread-ai/mxbai-rerank-xsmall-v1.
- https://huggingface.co/api/models/mixedbread-ai/mxbai-rerank-large-v2 recorded 2026-09-11
downloads: 102171; gated: false; private: false for mixedbread-ai/mxbai-rerank-large-v2.
Capability
4 medium confidenceRung 4, competitive frontier: real multilingual and code coverage and a widely used inference package, but the v2 generation shipped in March 2025 and its own reported BEIR average sits below the 2025-26 frontier. Level with jina-reranker, whose v3 card reports 61.94 BEIR. The two BEIR figures are not from one harness - jina own table credits mxbai-rerank-large-v2 with 61.44 - which is why this is placed by relation rather than by arithmetic on the two numbers, and why confidence is medium.
- https://huggingface.co/mixedbread-ai/mxbai-rerank-large-v2/raw/main/README.md recorded 2026-09-11
Benchmark Results table gives mxbai-rerank-large-v2 57.49 BEIR average, 29.79 multilingual, 84.16 Chinese, 32.05 code search and 0.89s latency, against 55.57 and 49.32 BEIR for base-v2 and large-v1.
- https://huggingface.co/jinaai/jina-reranker-v3/raw/main/README.md recorded 2026-09-11
Third-party table on the jina-reranker-v3 card places mxbai-rerank-large-v2 at 61.44 BEIR and mxbai-rerank-base-v2 at 58.40, below jina-reranker-v3 at 61.94.
Verified 2026-09-11