voyage
**Voyage AI: Domain-Specific Embeddings**
Voyage AI provides specialized embedding models optimized for specific domains (Finance, Code, Law) and retrieval tasks. While OpenAI's embeddings are "general purpose," Voyage models often outperform them on retrieval benchmarks (MTEB) due to specialized training.
**Key Models**
- **voyage-large-2**: High performance general purpose.
- **voyage-code-2**: Optimized for code retrieval (RAG on codebases).
- **voyage-finance-2**: Trained on financial documents (10-K, earnings calls).
- **voyage-law-2**: Optimized for legal contracts and case law.
**Context Length**
Voyage supports varying context lengths, often significantly larger than competitors, allowing for embedding entire documents rather than just chunks.
**Usage (Python)**
```python
import voyageai
vo = voyageai.Client(api_key="VOYAGE_API_KEY")
embeddings = vo.embed(
texts=["The court ruled."],
model="voyage-law-2",
input_type="document"
)
```
**Pricing**
Targeting enterprise users who need higher accuracy (Recall@K) to reduce hallucinations in RAG systems.