568M- Context
- 8Ktokens
- Released
- 2y ago
- Size
- 9.2Bparameters
BGE Multilingual Gemma2 is an open-weights multilingual embedding model from BAAI, part of its BGE (BAAI general embedding) line and trained on the google/gemma-2-9b base model. BAAI reports that it leads multilingual benchmarks such as MIRACL, MTEB-pl and MTEB-fr. It is built for embedding work including retrieval, classification and clustering.
How it compares
Additional source evidence
- Source model
- baai/bge-multilingual-gemma2
- Source family
- bge
- Source type
- embedding
- Last updated
- 2024-07-25
- Model context
- 8K
- Model output limit
- —
- Open weights
- Yes
- File attachments
- No
- Reasoning
- No
- Tool calling
- No
- Temperature control
- No
Downloadable weights
Complete source record (JSON)
{
"id": "baai/bge-multilingual-gemma2",
"name": "BGE Multilingual Gemma2",
"description": "Multilingual Gemma 2 based embedding model for retrieval and semantic search",
"family": "bge",
"type": "embedding",
"last_updated": "2024-07-25",
"attachment": false,
"reasoning": false,
"tool_call": false,
"temperature": false,
"modalities": {
"input": [
"text"
],
"output": [
"text"
]
},
"limit": {
"context": 8192,
"output": 0
},
"release_date": "2024-07-25",
"open_weights": true,
"license": "Gemma",
"weights": [
{
"label": "Hugging Face",
"url": "https://huggingface.co/BAAI/bge-multilingual-gemma2"
}
]
}Frequently asked questions
Running locally
Estimated VRAM is 6.9 GB at Q4, 11.8 GB at Q8 and 21.8 GB at FP16 for the 9.2B parameter model. These are estimates from the parameter count, covering weights plus about 15% and 0.5 GB for a short context.
Yes. A 24 GB GPU fits at FP16 (estimated 21.8 GB), 16 GB and 12 GB GPUs fit at Q8 (estimated 11.8 GB), and an 8 GB GPU fits at Q4 (estimated 6.9 GB).
Yes, at the estimated Q4 requirement of 6.9 GB. Assuming macOS lets the GPU use about 75% of unified memory, the smallest common Mac memory size that fits is 16 GB.
Privacy and data
Yes. BGE Multilingual Gemma2 has open weights, so you can self-host it for full control of your data.
Capabilities and specs
The context window is 8K tokens (8,192). A separate output length is not listed.
No. The input modality is text only, so it does not read images, audio, video or PDFs.
No, tool calling (function calling) is not supported.
No, a reasoning or thinking mode is not supported.
Quality and benchmarks
The closest alternatives are Gemma 2 2B IT by Google DeepMind, DCLM 7B by Apple and OLMoE 1B-7B by Ai2. They are matched on model kind and release date, with no comparable price available.
Gemma 2 2B ITBGE Multilingual Gemma2 vs Gemma 2 2B ITDCLM 7BBGE Multilingual Gemma2 vs DCLM 7BOLMoE 1B-7BBGE Multilingual Gemma2 vs OLMoE 1B-7B
Licensing and access
BGE Multilingual Gemma2 is open weights, under the Gemma license.
The weights can be downloaded from the Hugging Face repository BAAI/bge-multilingual-gemma2.
Basics
BAAI made BGE Multilingual Gemma2, and it was released on July 25, 2024.
Yes. The model is open weights, so the weights are available to download and use.
Generated from Vertical Tabs data, updated November 2, 2026.
More in Bge
568M567M
Discussion
No comments yet
Be the first to share your thoughts.