Jina Reranker m0
Multilingual multimodal reranker for visually rich documents.
About
Jina Reranker m0 is a 2.4B-parameter reranker built on Qwen2-VL-2B with LoRA tuning. Unlike classic cross-encoder rerankers, it accepts images directly, so it can rank pages of mixed text, figures, and tables across 29+ languages with up to 10K tokens of context. Weights are available under CC-BY-NC 4.0, which limits them to research; commercial use requires Jina's API.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Team-reported: BEIR 58.95, MIRACL 66.75, MKQA 68.19
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Specs (2.4B, Qwen2-VL-2B + LoRA, 10240 tokens, 29+ languages, CC-BY-NC-4.0) verified via HF model card, PromptLayer, and GPUStack PR.; Benchmark figures (BEIR 58.95, MIRACL 66.75) taken from Jina's own reranker-v3 comparison table; team-reported.; Exact release day not verified; commercial terms not fully documented; scored one point lower for thinner verification..