by Sentence TransformersReleased 18 November 2025
Multi-qa-mpnet-base-dot-v1 turns questions and passages into fingerprints of their meaning, so a question finds the paragraph that answers it. It was trained on about 215 million question and answer pairs, takes inputs of up to about half a page, and its makers advise comparing fingerprints with a dot product. Pick it for question-answering search. For general similarity or grouping, use all-mpnet-base-v2.
You pay for what each request uses. No subscription, no minimum.
One page, the ordinary use
Every rate
| Meter | ↑ you send | ↓ comes back |
|---|---|---|
| Text tokens | 1,66 DA |
per 1M tokens
One company runs this model. OpenDunes sends every request to it. DA per 1M tokens: ↑ what you send.
You pay the price of the provider that served your request. Prices read 15:19 · speed over the last 30 minutes · uptime over 3 days.
Swap the base URL and the key. Your OpenAI code runs unchanged.
import os
from openai import OpenAI
client = OpenAI(
base_url="https://opendunes.com/v1",
api_key=os.environ["OPENDUNES_API_KEY"],
)
resp = client.embeddings.create(
model="sentence-transformers/multi-qa-mpnet-base-dot-v1",
input="Delivery to Oran takes two days.",
)
print(resp.data[0].embedding[:5])Runs this request once on your balance.
About 0,0017 DA per page. ↑ 1,66 DA per 1M tokens.
No. You pay per use from your balance, with no subscription. Run it once from Call it in code, at the same price.
Multi-qa-mpnet-base-dot-v1 turns questions and passages into fingerprints of their meaning, so a question finds the paragraph that answers it. It was trained on about 215 million question and answer pairs, takes inputs of up to about half a page, and its makers advise comparing fingerprints with a dot product. Pick it for question-answering search. For general similarity or grouping, use all-mpnet-base-v2.
Deposit dinars with your Edahabia or CIB card. Every model, multi-qa-mpnet-base-dot-v1 included, is paid from that balance. No foreign card, no conversion.
The model card tags English only. Arabic, French and Darja are not named.
Up to 512 tokens. Split anything longer.
768
Sentence Transformers. Released 18 November 2025.
Send All-mpnet-base-v2 a sentence or a short passage and it returns a fingerprint of what the text means, so a search can find text that says the same thing in different words. Use it for finding documents, grouping similar texts, scoring closeness and ranking. Its makers rate it the best quality among their general models, and the smaller all-MiniLM-L6-v2 is about five times faster.
Send All-MiniLM-L6-v2 a sentence or a short passage and it returns a fingerprint of what the text means, so a search can find text that says the same thing in different words. You store the fingerprints and compare them for search, grouping similar texts and ranking. Its makers describe it as about five times faster than all-mpnet-base-v2, with good quality.
Send All-MiniLM-L12-v2 a sentence or a short passage and it returns a fingerprint of what the text means, so a search can find text that says the same thing in different words. Use it for semantic search, grouping similar texts and scoring how close two texts are. Its sources give it the same uses as all-MiniLM-L6-v2, so test both on a sample of your own texts.
Send Paraphrase-MiniLM-L6-v2 a sentence or a short passage and it returns a fingerprint of what the text means. Its strength is spotting paraphrases, two sentences that say the same thing in different words, which helps in finding duplicates, scoring closeness and grouping. It is small and light.
From what its maker and two benchmark sites publish, and where it stands on OpenDunes. Read 7 Nov 2026.
Open weights
The model's weights are public.
Find FAQ answers that match a customer's question
Group similar feedback messages
Spot duplicate product titles