Reranking / MODEL API
BGE-Reranker-V2-M3
Input $0.03 / 1M tokens
OVERVIEW
Model overview
bge-reranker-v2-m3，Different from embedding model, reranker uses question and document as input and directly output similarity instead of embedding. You can get a relevance score by inputting query and passage to the reranker. And the score can be mapped to a float value in [0,1] by sigmoid function.
Specifications
- Category
- Reranking
- Input modalities
- text
- Output modalities
- text
MODEL PRICING
Model API pricing
Reranking
USD per 1M tokens| Model | Input | Output |
|---|---|---|
| BGE-Reranker-V2-M3 | $0.03 | — |
Public platform rates refresh every five minutes. A dash means a rate is not provided; it does not mean free. Check the console for final charges.
BUILT FOR YOUR STACK
API quickstart
Expand connection setup and run the preparation commands first. Then save and run the Python or cURL example.
Connection setup & preparationActual API address, model ID, and environment variables
- API address
https://api.ecohash.com/v1- API model ID
bge-reranker-v2-m3
Replace YOUR_API_KEY with the full inference API key issued in the console. Run setup in macOS/Linux or a compatible shell. JSON cURL examples use the Python standard library to build the request body.
python -m pip install requests
export GIGACOMMA_API_BASE_URL='https://api.ecohash.com/v1'
export GIGACOMMA_MODEL_ID='bge-reranker-v2-m3'
export GIGACOMMA_API_KEY="YOUR_API_KEY"GIGACOMMA_API_KEY is the environment variable name. Its value is an inference API key from the console, not an OpenAI API key. Get an API key
import osimport requestsMODEL_ID = os.environ["GIGACOMMA_MODEL_ID"]BASE_URL = os.environ["GIGACOMMA_API_BASE_URL"].rstrip("/")headers = {"Authorization": f"Bearer {os.environ['GIGACOMMA_API_KEY']}"}response = requests.post( f"{BASE_URL}/rerank", headers=headers, json={"model": os.environ["GIGACOMMA_MODEL_ID"], "query": "What is GPU computing?", "documents": ["GPUs accelerate parallel computing.", "A guide to baking bread."], "top_n": 2}, timeout=120,)response.raise_for_status()print(response.json())