Embeddings with Velqa.dev

Velqa offers two multilingual embedding models (strong in French and Arabic), ideal for semantic search, RAG and document comparison. They are available on the OpenAI-compatible API https://api.velqa.dev/v1/embeddings and included in every plan (Starter, Dev, Pro, Recharge Boost).

• bge-m3 — fast and cheap, 1024 dimensions, 8k context. A solid default.
• qwen3-embedding-8b — higher quality (top multilingual MTEB), up to 4096 dimensions, 32k context. Prefer it for demanding RAG.

Pricing

Embeddings produce no output tokens: usage is charged on the input tokens sent to the model. Current USD pricing and checkout availability are shown in the dashboard; checkout is Stripe, in USD, by international card.

Example with curl

curl https://api.velqa.dev/v1/embeddings \
  -H "Authorization: Bearer $VELQA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "bge-m3", "input": ["First document", "ثاني وثيقة"]}'

Example with Python

from openai import OpenAI

client = OpenAI(base_url="https://api.velqa.dev/v1", api_key="VELQA_API_KEY")
resp = client.embeddings.create(
    model="bge-m3",
    input=["First document", "ثاني وثيقة"]
)
print(len(resp.data[0].embedding))

The response contains one float vector per input item, plus usage.prompt_tokens telling you how many tokens were billed.

See also

• RAG pipeline — the full guide: index, search, rerank, answer.
• Reranking — reorder candidates before sending them to the chat model.