mxbai-rerank-large-v2
mxbai-rerank-large-v2 is a 1.54B-parameter listwise reranker from Mixedbread AI built on Qwen2.5-1.5B, supporting 100+ languages and a 32K-token context with native code and instruction-following retrieval awareness. Published Apache-2.0 in 2026, it reports BEIR ~57.8 and outperforms bge-reranker-v2-m3 on English nDCG@10 while remaining permissively licensed.
The current open-weight quality leader in reranking, riding on Qwen2.5-1.5B's strong base. If you can afford the 1.5B inference cost, it outscores bge-reranker-v2-m3 on English benchmarks at comparable multilingual coverage. Default to bge-reranker-v2-m3 for cost, upgrade to this when accuracy matters.
Overview
mxbai-rerank-large-v2 is a 1.54B-parameter listwise reranker from Mixedbread AI built on Qwen2.5-1.5B, supporting 100+ languages and a 32K-token context with native code and instruction-following retrieval awareness. Published Apache-2.0 in 2026, it reports BEIR ~57.8 and outperforms bge-reranker-v2-m3 on English nDCG@10 while remaining permissively licensed.
Strengths
- Qwen2.5-1.5B backbone gives 32K context and native instruction-following at rerank time
- Reports BEIR nDCG@10 ~57.8 — top-of-class for open rerankers in 2026
- Apache-2.0 weights — clean commercial licensing unlike Jina v2
- Strong on code and structured retrieval (inherits Qwen2.5 pretraining)
Weaknesses
- 1.54B params — 3x larger than bge-reranker-v2-m3 at similar multilingual quality
- Decoder architecture with last-token scoring is slower per-pair than encoder cross-encoders
- Model card notes commercial use above certain revenue thresholds requires Mixedbread contact despite Apache-2.0 weights — read terms carefully
- Requires Qwen2-compatible inference stack (transformers >=4.45, vLLM, or TEI)
Quantization variants
Each quantization trades model quality for file size and VRAM. Q4_K_M is the most popular starting point.
| Quantization | File size | VRAM required |
|---|---|---|
| Q4_K_M | 0.8 GB | 2 GB |
Get the model
HuggingFace
Original weights
Source repository — direct quantization required.
Hardware that runs this
Cards with enough VRAM for at least one quantization of mxbai-rerank-large-v2.
Models worth comparing
Same parameter band, plus what's one tier above and below — so you can decide what actually fits your hardware.
Frequently asked
What's the minimum VRAM to run mxbai-rerank-large-v2?
Can I use mxbai-rerank-large-v2 commercially?
What's the context length of mxbai-rerank-large-v2?
Source: huggingface.co/mixedbread-ai/mxbai-rerank-large-v2
Reviewed by RunLocalAI Editorial. See our editorial policy for how we research and verify model claims.
Related — keep moving
Verify mxbai-rerank-large-v2 runs on your specific hardware before committing money.