Skip to main content
Watch: I Built an AI App Without Leaving Databricks

BAAI rerankers - verified

BAAI reranker models

BAAI (open weights): 5 reranker models, read from the provider's own pages on October 18, 2026, 5 with open weights. BAAI rerankers are self-hosted from open weights, with no per-call price.

Read from huggingface.co. Compare every reranker

Models
5
Current
5
Older
0
Open weights
5

Billing and caveats

BAAI (open weights)

Languages come from the model list on the bge-reranker-v2-m3 card. Status GA means released on Hugging Face (inferred).

Models

bge-reranker-v2-m3

BAAI/bge-reranker-v2-m3 - Open weights - Self-host (open weights or NIM containers)

GAinferred
Price (provider's unit)
Not stated

No per-call price: self-hosted.

Status

GAinferred

Status inferred: the source lists the model but does not label it.

Context
Not stated
Documents per request
Not stated
Languages

Multilingual (no count stated)

“[BAAI/bge-reranker-v2-m3](https://huggingface.co/BAAI/bge-reranker-v2-m3) · [bge-m3](https://huggingface.co/BAAI/bge-m3) · Multilingual”
huggingface.co
Open weights

Yes - Apache 2.0, 568M parameters

BAAI/bge-reranker-v2-m3 - license

“"license":"apache-2.0"”
huggingface.co
“"total":567755777”
567,755,777 parameters - huggingface.co

Built on bge-m3. The card prose states no max length (only code samples set one), so maxContext is null.

bge-reranker-v2-gemma

BAAI/bge-reranker-v2-gemma - Open weights - Self-host (open weights or NIM containers)

GAinferred
Price (provider's unit)
Not stated

No per-call price: self-hosted.

Status

GAinferred

Status inferred: the source lists the model but does not label it.

Context
Not stated
Documents per request
Not stated
Languages

Multilingual (no count stated)

“[BAAI/bge-reranker-v2-gemma](https://huggingface.co/BAAI/bge-reranker-v2-gemma) · [gemma-2b](https://huggingface.co/google/gemma-2b) · Multilingual”
huggingface.co
Open weights

Yes - Apache 2.0, 2.51B parameters

BAAI/bge-reranker-v2-gemma - license

“"license":"apache-2.0"”
huggingface.co
“"total":2506172416”
2,506,172,416 parameters - huggingface.co

LLM-based reranker initialised from google/gemma-2b; the card metadata says apache-2.0, but Gemma base-model terms may also apply.

bge-reranker-v2-minicpm-layerwise

BAAI/bge-reranker-v2-minicpm-layerwise - Open weights - Self-host (open weights or NIM containers)

GAinferred
Price (provider's unit)
Not stated

No per-call price: self-hosted.

Status

GAinferred

Status inferred: the source lists the model but does not label it.

Context
Not stated
Documents per request
Not stated
Languages

Multilingual (no count stated)

“[BAAI/bge-reranker-v2-minicpm-layerwise](https://huggingface.co/BAAI/bge-reranker-v2-minicpm-layerwise) · [MiniCPM-2B-dpo-bf16](https://huggingface.co/openbmb/MiniCPM-2B-dpo-bf16) · Multilingual · 8-40”
huggingface.co
Open weights

Yes - Apache 2.0, 2.72B parameters

BAAI/bge-reranker-v2-minicpm-layerwise - license

“"license":"apache-2.0"”
huggingface.co
“"total":2724956928”
2,724,956,928 parameters - huggingface.co

Layerwise model: output layers 8-40 can be selected to trade accuracy for speed.

bge-reranker-large

BAAI/bge-reranker-large - Open weights - Self-host (open weights or NIM containers)

GAinferred
Price (provider's unit)
Not stated

No per-call price: self-hosted.

Status

GAinferred

Status inferred: the source lists the model but does not label it.

Context
Not stated
Documents per request
Not stated
Languages

Chinese and English

“[BAAI/bge-reranker-large](https://huggingface.co/BAAI/bge-reranker-large) · [xlm-roberta-large](https://huggingface.co/FacebookAI/xlm-roberta-large) · Chinese and English”
huggingface.co
Open weights

Yes - MIT, 560M parameters

BAAI/bge-reranker-large - license

“"license":"mit"”
huggingface.co
“"total":559891971”
559,891,971 parameters - huggingface.co

First-generation cross-encoder on xlm-roberta-large.

bge-reranker-base

BAAI/bge-reranker-base - Open weights - Self-host (open weights or NIM containers)

GAinferred
Price (provider's unit)
Not stated

No per-call price: self-hosted.

Status

GAinferred

Status inferred: the source lists the model but does not label it.

Context
Not stated
Documents per request
Not stated
Languages

Chinese and English

“[BAAI/bge-reranker-base](https://huggingface.co/BAAI/bge-reranker-base) · [xlm-roberta-base](https://huggingface.co/xlm-roberta-base) · Chinese and English”
huggingface.co
Open weights

Yes - MIT, 278M parameters

BAAI/bge-reranker-base - license

“"license":"mit"”
huggingface.co
“"total":278044931”
278,044,931 parameters - huggingface.co

First-generation cross-encoder on xlm-roberta-base.

FAQ

How much does BAAI reranking cost?

bge-reranker-v2-m3: no per-call price (self-hosted). bge-reranker-v2-gemma: no per-call price (self-hosted). bge-reranker-v2-minicpm-layerwise: no per-call price (self-hosted). bge-reranker-large: no per-call price (self-hosted). bge-reranker-base: no per-call price (self-hosted). Prices are in BAAI's own units as of October 18, 2026.

What is the context length of BAAI rerankers?

BAAI's pages do not state a context length for these models.

Can I run BAAI rerankers myself?

Yes for bge-reranker-v2-m3 (Apache 2.0), bge-reranker-v2-gemma (Apache 2.0), bge-reranker-v2-minicpm-layerwise (Apache 2.0), bge-reranker-large (MIT) and bge-reranker-base (MIT): the weights are on Hugging Face.

Sources

Last verified from each provider's own docs, pricing pages and Hugging Face cards; every quote was string-matched against the page that day. Prices are shown in the provider's own unit and change without notice, so check the linked source before you commit a budget.

Newsletter

Know when a reranker API changes

One email per week. New model releases, retrieval and pricing updates and what they mean for what you build.