e5-mistral-7b-instruct
text-embeddingintfloat · intfloat · released
About
E5-Mistral-7B-Instruct is intfloat's open-weight text embedding model built on a Mistral-7B backbone. It produces high-quality dense embeddings for retrieval, RAG and semantic search, and is a strong open baseline on MTEB.
Architecture
- Type
- encoder
- Parameters
- 7.1B
E5-Mistral text embedding model: a Mistral-7B backbone (MistralModel) fine-tuned as a dense text encoder producing sentence/passage embeddings. 7.11B parameters. One of the strongest open-weight embedding models on MTEB-style retrieval benchmarks.
Memory
- Weights (BF16)
- 14.22 GB
- Activation estimate
- 2.00 GB
Pricing
Free — open weights
Self-host on your own GPU. The calculator surfaces GPU-hours cost on the hardware page instead of an API price.
Provenance
- Source
- huggingface.co
- License
- mit
- Hugging Face
- intfloat/e5-mistral-7b-instruct
- Last verified
- 2026-06-24
embeddingretrievalragtext-embeddingopen-weight