Embeddings Endpoints

IG1 AI provides two compatible embedding endpoints:

Compatibility	Endpoint	Models
OpenAI	POST /v1/embeddings	bge-m3, qwen3-vl-embedding-8b
Cohere	POST /cohere/v2/embed	bge-m3, qwen3-vl-embedding-8b

---

Which Model should I use?

Simple and Fast text embeddings:
→ Use bge-m3 model with OpenAI or Cohere's endpoints.

Multimodal inputs (text + image), dynamic dimensions, or query/doc differentiation
→ Use the Cohere-compatible endpoint with qwen3-vl-embedding-8b model.

info

• Model dimensions are fixed to 1024 when using bge-m3 model.
• "input_type": "image" is not supported by bge-m3 model.