Skip to content
English

Switch theme

// BGE embeddings & rerankers · Reranker

bge-reranker-v2-m3: VRAM & local deployment

568M parameters. About 0.9 GB at 4-bit and 1.7 GB at BF16.

released 2024-03-15licence: Commercial OKmemory figures are estimates

01Specs

PublisherBAAI · BAAI
Family / seriesBGE embeddings & rerankers · bge-reranker-m3
TypeReranker (text-classification)
Parameters568M Source: counted from official safetensors weights
Released weightsFP32 · 2.3 GBdownload
LicenceApache-2.0 · Commercial OKPermissive licence (Apache-2.0 / MIT / CC-BY / OpenMDW): commercial use allowed.
Released2024-03-15 (Hugging Face repo created)

02How much VRAM it needs (estimate)

Rerankers are small; memory grows with batch size and sequence length.

PrecisionWeightsWeights + overheadSmallest GPUMac
BF16 / FP16 (original)1.1 GB 1.7 GB 8GB GPUMac 16GB
FP8 / INT8 / Q8_00.6 GB 1.2 GB 8GB GPUMac 16GB
4-bit (Q4_K_M / AWQ / GPTQ-Int4)0.3 GB 0.9 GB 8GB GPUMac 16GB

Bits per weight: BF16 16, FP8 8.5, 4-bit 4.85. “Fits” = total ≤ 90% of memory; Mac GPU can use ~2/3 of unified memory (≤36GB) or ~3/4 (larger). For MoE, all experts must be in memory, so size by total parameters. Method →

03How to run it

Only runtimes the publisher or the runtime’s official library names for this model.

Which inference engine to choose →

04Official repositories & quantised builds

1 official repo(s) from BAAI.

05Other BGE embeddings & rerankers models

All 8 →
ModelParams4-bit totalReleased
bge-code-v11.54B1.5 GB2025-05-15
BGE-VL-base150M0.6 GB2025-02-25
BGE-VL-large428M0.8 GB2025-02-25
bge-en-icl7.11B5.2 GB2024-07-25
bge-multilingual-gemma29.24B6.7 GB2024-07-25
bge-reranker-v2.5-gemma2-lightweight9.24B6.7 GB2024-07-25
bge-m3——2024-01-27

06Sources

Data checked 2024-06-24. Parameters, context and licence are copied from the official repository; memory figures are our estimates.

Data updated: Self-host open models(UTC+8)