BAAI/bge-m3
BAAI/bge-m3
BGE-M3, multilingual embedding, 8192 context, 1024 dim.
Parameters
558M
Context
8.2K
Downloads
34.9M
Likes
3.4K
Architecture
xlm-roberta
24 layers · head 64 · hidden 1024
GPU Requirements
VRAM / RAM / disk estimates by quantization and usage scenario
| Quantization | Minimum | Recommended | Production |
|---|---|---|---|
| FP1616bit | 2.7 GB RAM 32.0 · Disk 40.7 | 3.0 GB RAM 32.0 · Disk 66.7 | 3.3 GB RAM 32.0 · Disk 105.7 |
| FP88bit | 2.0 GB RAM 32.0 · Disk 39.8 | 2.2 GB RAM 32.0 · Disk 65.8 | 2.5 GB RAM 32.0 · Disk 104.8 |
| INT88bit | 2.0 GB RAM 32.0 · Disk 39.8 | 2.2 GB RAM 32.0 · Disk 65.8 | 2.5 GB RAM 32.0 · Disk 104.8 |
| AWQ4bit | 1.6 GB RAM 32.0 · Disk 39.4 | 1.8 GB RAM 32.0 · Disk 65.4 | 2.1 GB RAM 32.0 · Disk 104.4 |
| GPTQ4bit | 1.6 GB RAM 32.0 · Disk 39.4 | 1.8 GB RAM 32.0 · Disk 65.4 | 2.1 GB RAM 32.0 · Disk 104.4 |
| GGUF4bit | 1.6 GB RAM 32.0 · Disk 39.4 | 1.8 GB RAM 32.0 · Disk 65.4 | 2.1 GB RAM 32.0 · Disk 104.4 |
GPU Recommendations
1× RTX 5060 Ti
8 GB VRAM
≈ 3319 vec/s
Effective production throughput ≈ 2655.2 vectors/s
score 0.552
1× RTX 5060 Ti
8 GB VRAM
≈ 3319 vec/s
Effective production throughput ≈ 2655.2 vectors/s
score 0.552
8× RTX 5060 Ti
64 GB VRAM
≈ 26548 vec/s
Effective production throughput ≈ 21238.4 vectors/s
score 0.3636
2× RTX 5060 Ti
16 GB VRAM
≈ 6637 vec/s
Effective production throughput ≈ 5309.6 vectors/s
score 0.451
💬 Community — real deployment experience
0 Articles · 0 Benchmarks
GPUs that fit (single card)
Tesla V100
16 GB VRAM
$21.1
/mo · vast
RTX 5060 Ti
8 GB VRAM
$39.7
/mo · vast
RTX 4070S Ti
16 GB VRAM
$49.5
/mo · vast
RTX 5070
12 GB VRAM
$58.7
/mo · vast
RTX 5070 Ti
16 GB VRAM
$62.0
/mo · vast