All tasks / Models
bge-m3
BAAI's MIT-licensed multilingual embedding model with dense, sparse and multi-vector retrieval in one.
- BAAI
- Open weights
- Released Jan 2024
Where it ranks
- –Embeddings & retrievalModels · Local · no result on this task’s benchmarks yet
Facts
- Vendor
- BAAI
- Family
- BGE-M3
- Released
- Jan 2024
- Weights
- Open: download and run it yourself
- License
- MIT
- Parameters
- 0.57B
- GPU memory to run locally
- ~3 GB (Board lists ~2.2 GB memory use (0.57B parameters, fp32).)
- Access
- Free (open weights)
- Last verified
- 19 Oct 2026
Benchmark results
| Benchmark | Result | Entry on the leaderboard | As of | Source |
|---|---|---|---|---|
| Agentset Embedding Model Leaderboard | 1,480 | BAAI/bge-m3ELO from GPT-5-judged pairwise comparisons; rank 13 of 18; win rate 44.3%; page says 'Last updated: February 15, 2026'. | 15 Feb 2026 | agentset.ai |
Where you can use it
- SOllamaVector databases & engines · Local
- SOpenRouterVector databases & engines · Cloud
- SvLLMVector databases & engines · Local