E5-Mistral

E5-Mistral#

E5-Mistral-7B-Instruct is an instruction-aware text-embedding model built by fine-tuning the dense Mistral-7B-v0.1 decoder backbone. It produces 4,096-dimensional dense embeddings through last-token pooling and L2 normalization for retrieval and semantic similarity tasks. Although it was fine-tuned on multilingual data, the upstream model card recommends it primarily for English.

FuriosaAI publishes E5-Mistral-7B-Instruct under the furiosa-ai organization on the Hugging Face Hub, shipping a Furiosa Executable Bundle (FXB) for running it on FuriosaAI RNGD with Furiosa-LLM. The same upstream weights also run on other frameworks (such as Sentence Transformers and Transformers); for usage with those, see the upstream model card linked below.

For other instruction-aware embedding models, see Harrier OSS v1 and Qwen3-Embedding.

Available Models#

Model

Quantization

RNGD cards

Notes

furiosa-ai/e5-mistral-7b-instruct

None (FP16)

1

7B instruction-aware text embedding

  • Architecture: Mistral (dense decoder backbone), MistralModel

  • Task: Embedding

  • Input / Output: Text / Dense embeddings (vector)

  • Quantization: No quantization — the model uses its upstream FP16 weights.

Usage#

To run this model with Furiosa-LLM, follow the examples below after installing Furiosa-LLM and its prerequisites. You can use the model either online through the OpenAI-compatible server or offline through the Furiosa-LLM Python API.

Launch the server#

Serve the model by passing its furiosa-ai/<repo> identifier:

# Launch the server, listening on port 8000 by default
furiosa-llm serve furiosa-ai/e5-mistral-7b-instruct

When the server is ready, you will see:

INFO:     Started server process [27507]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)

Basic Usage#

The server exposes an OpenAI-compatible /v1/embeddings endpoint. E5-Mistral is instruction-aware: prepend a one-sentence task description to each query in the Instruct: ...\nQuery: ... format, and do not add the instruction to documents. For more details, see the upstream model card. Request embeddings with curl:

curl http://localhost:8000/v1/embeddings \
    -H "Content-Type: application/json" \
    -d '{
    "model": "furiosa-ai/e5-mistral-7b-instruct",
    "input": [
      "Instruct: Given a web search query, retrieve relevant passages that answer the query\nQuery: summit define",
      "Definition of summit: the highest point of a mountain."
    ]
    }' \
    | python -m json.tool

Because the endpoint is OpenAI-compatible, you can also use the OpenAI Python client:

from openai import OpenAI

client = OpenAI(base_url="http://localhost:8000/v1", api_key="EMPTY")

query = (
    "Instruct: Given a web search query, retrieve relevant passages that answer the query\n"
    "Query: summit define"
)
document = "Definition of summit: the highest point of a mountain."

response = client.embeddings.create(
    model="furiosa-ai/e5-mistral-7b-instruct",
    input=[query, document],
)

for data in response.data:
    print(f"Index {data.index}: {len(data.embedding)} dimensions")

Advanced Usage#

For offline use, load the model with the LLM constructor (the FXB shipped in the repo is discovered automatically) and call embed to obtain L2-normalized dense vectors. Their dot product is therefore the cosine similarity:

from furiosa_llm import LLM

query = (
    "Instruct: Given a web search query, retrieve relevant passages that answer the query\n"
    "Query: summit define"
)
document = "Definition of summit: the highest point of a mountain."

with LLM("furiosa-ai/e5-mistral-7b-instruct") as llm:
    outputs = llm.embed([query, document])
    embeddings = [output.outputs.embedding for output in outputs]

similarity = sum(a * b for a, b in zip(*embeddings, strict=True))
print(f"Cosine similarity: {similarity:.4f}")

Learn more#