bge reranker v2 m3

An open model that scores how well each passage fits a query. BAAI/bge-reranker-v2-m3

Get started Run in your cloud Hugging Face

Accuracy

0.676

AskUbuntu, nDCG@10
Time per request

Not timed one request at a time
Serverless price

Run in your cloud

Call it

from sie_sdk import SIEClient
client = SIEClient("http://localhost:8080")
query = "queryWhy does my model take 30 seconds to answer the first request?"
documents = [
    "document 1Cold storage keeps old checkpoints cheap to retain.",
    "document 2Subsequent requests are fast once the model is warm.",
    "document 3The first request is slow because the model loads into GPU memory; keep one replica warm.",
]
items = [{"id": str(i), "text": d} for i, d in enumerate(documents)]
ranked = client.score("modelBAAI/bge-reranker-v2-m3", {"text": query}, items)
for r in ranked["scores"]:
    print(r["score"], documents[int(r["item_id"])])

import { SIEClient } from '@superlinked/sie-sdk';

const client = new SIEClient('http://localhost:8080');

const query = "Why does my model take 30 seconds to answer the first request?";
const documents = [
  "Cold storage keeps old checkpoints cheap to retain.",
  "Subsequent requests are fast once the model is warm.",
  "The first request is slow because the model loads into GPU memory; keep one replica warm.",
];
const items = documents.map((text, i) => ({ id: String(i), text }));
const ranked = await client.score('BAAI/bge-reranker-v2-m3', { text: query }, items);
console.log(ranked.scores.map(({ itemId, score }) => ({ document: documents[Number(itemId)], score })));

curl http://localhost:8080/v1/score/BAAI%2Fbge-reranker-v2-m3 \
  -H "Content-Type: application/json" \
  -d "{\"query\":{\"text\":\"Why does my model take 30 seconds to answer the first request?\"},\"items\":[{\"id\":\"0\",\"text\":\"Cold storage keeps old checkpoints cheap to retain.\"},{\"id\":\"1\",\"text\":\"Subsequent requests are fast once the model is warm.\"},{\"id\":\"2\",\"text\":\"The first request is slow because the model loads into GPU memory; keep one replica warm.\"}]}"

Build the "Rerank" capability into my app using the Superlinked Inference Engine (SIE).

Context
- SIE is an OpenAI-style inference API. Python SDK: `from sie_sdk import SIEClient`; TypeScript: `@superlinked/sie-sdk`.
- Base URL: http://localhost:8080, my own SIE server, which takes no API key.
- Model: BAAI/bge-reranker-v2-m3 (SIE primitive: /score). Keep the model id configurable.

Task
- Input: a query string plus a list of candidate documents.
- Behaviour: reorder the candidates by true relevance to the query using the reranking model
- Call the selected SIE primitive once per request and map the response into your domain type.

Deliverables
- A typed client wrapper, an application-level function for this task, error handling for timeouts/empty input, and unit tests with a stubbed client.
- Wire it into my existing stack (ask me which framework if unclear) and add a short usage example.

Run

Next to the other rerankers

Next to the other rerankers
Model	AskUbuntu, nDCG@10	Time per request	Serverless price	Size
bge reranker v2 m3 BAAI	0.676	Not timed	Run in your cloud	568M
Qwen3 Reranker 4B Qwen	0.697	55 ms per query	$0.0425 per 1M tokens	4.0B
mxbai rerank large v2 mixedbread-ai	0.691	Not timed	Run in your cloud	1.5B
gte reranker modernbert base Alibaba-NLP	0.670	Not timed	Run in your cloud	150M
bge m3 BAAI	0.666	Not timed	$0.0743 per 1M tokens $3.72 per 1M pairs	568M

All 22 rerankers

Measured on public datasets

Measured on public datasets
Dataset	Metric	Value
CMedQAv2Reranking	MAP@10	0.829
CMedQAv1Reranking	MAP@10	0.819
LegalBenchConsumerContractsQA	nDCG@10	0.816
SciFact	nDCG@10	0.746
AskUbuntuDupQuestions	nDCG@10	0.676
StackOverflowQA	nDCG@10	0.616
FiQA2018	nDCG@10	0.450
CQADupstackPhysicsRetrieval	nDCG@10	0.417
CosQA	nDCG@10	0.290
SCIDOCS	nDCG@10	0.187

Facts

Size: 568M parameters
License: Apache-2.0
Reads: Text, up to 8,192 tokens
Endpoint: score

Evaluation

Accuracy: Run on SIE against the public dataset named in each figure
Reranking: Reorders the top 50 from gte multilingual base
Time: SIE timed on a GPU server with the client alongside, one request at a time
Prices: Serverless list prices

Run bge reranker v2 m3 in your own cloud, or start on Serverless

SIE Serverless

We host the models; you call one API.

No idle GPUs, pay for what you use
Fits your stack: SDK, API, CLI, MCP
Self-host the same open-source engine
SOC 2 Type 2, US or EU data residency
Create an account to try Serverless

Get started

SIE Cluster

Kubernetes in your own cloud, backed by our engineers.

Terraform modules for your cloud
Apache-2.0, same engine as Serverless
GPU workers scale to zero between jobs
Separate worker pools for each tenant

Agent prompt Deploy SIE to our AWS account with the superlinked/sie/aws Terraform module. Docs: superlinked.com/docs/deployment.

Deploy SIE to our Google Cloud project with the superlinked/sie/google Terraform module. Docs: superlinked.com/docs/deployment.

Deploy SIE to our Azure subscription with the superlinked/sie/azure Terraform module. Docs: superlinked.com/docs/deployment.

Deploy SIE to our Alibaba Cloud account with the superlinked/sie/alicloud Terraform module. Docs: superlinked.com/docs/deployment.

Talk to us Deploy guide

SIE Local

The open-source server on your own machine.

Runs on your NVIDIA GPU
One container, no cluster
Process your data fully offline
Same SDK and model IDs as Serverless

docker run --gpus all -p 8080:8080 ghcr.io/superlinked/sie-server:latest-cuda12-default

Quickstart