other
7.11B parameters
Commercial OK
Reviewed May 2026

E5 Mistral 7B Instruct

E5-Mistral-7B-Instruct is a 7.11B-parameter decoder-based embedder fine-tuned from Mistral-7B by Microsoft's intfloat team, producing 4096-dim embeddings with the model's native 32K context. It uses task-conditioned instructions at the query side and sat at the top of MTEB on release with a score of ~66.6, anchoring the modern LLM-as-embedder paradigm.

License: mit·Context: 32,768 tokens
BLK · VERDICT

Our verdict

OP · Fredoline Eruo|VERIFIED MAY 29, 2026
unrated

The model that proved decoder LLMs make excellent embedders, and still a credible choice when quality dominates cost. But 4096-dim storage and 7B inference are real costs — modern 500M-class encoders deliver within 2 MTEB points at 1/10 the footprint. Reach for it only when the use case explicitly needs instruction-conditioning or 32K context.

Overview

E5-Mistral-7B-Instruct is a 7.11B-parameter decoder-based embedder fine-tuned from Mistral-7B by Microsoft's intfloat team, producing 4096-dim embeddings with the model's native 32K context. It uses task-conditioned instructions at the query side and sat at the top of MTEB on release with a score of ~66.6, anchoring the modern LLM-as-embedder paradigm.

Strengths

  • Top-tier MTEB English score (~66.6) — still competitive with 2025 LLM-class embedders
  • Instruction-conditioned: phrase the task in natural language at query time, no fine-tune required
  • 4096-dim vectors with 32K context — handle whole documents without chunking
  • MIT license, no usage restrictions

Weaknesses

  • 7B params + 4096-dim vectors — ~10x storage and ~30x inference cost vs. BGE-large for marginal MTEB gain
  • Requires GPU (~15GB VRAM fp16) — no realistic CPU deployment
  • Decoder-architecture with last-token pooling — slower per-passage than BERT-class encoders
  • Largely synthetic GPT-4 training data raises distillation/license concerns for downstream training

Quantization variants

Each quantization trades model quality for file size and VRAM. Q4_K_M is the most popular starting point.

QuantizationFile sizeVRAM required
Q4_K_M3.9 GB5 GB

Get the model

HuggingFace

Original weights

huggingface.co/intfloat/e5-mistral-7b-instruct

Source repository — direct quantization required.

Hardware that runs this

Cards with enough VRAM for at least one quantization of E5 Mistral 7B Instruct.

Compare alternatives

Models worth comparing

Same parameter band, plus what's one tier above and below — so you can decide what actually fits your hardware.

Frequently asked

What's the minimum VRAM to run E5 Mistral 7B Instruct?

5GB of VRAM is enough to run E5 Mistral 7B Instruct at the Q4_K_M quantization (file size 3.9 GB). Higher-quality quantizations need more.

Can I use E5 Mistral 7B Instruct commercially?

Yes — E5 Mistral 7B Instruct ships under the mit, which permits commercial use. Always read the license text before deployment.

What's the context length of E5 Mistral 7B Instruct?

E5 Mistral 7B Instruct supports a context window of 32,768 tokens (about 33K).

Source: huggingface.co/intfloat/e5-mistral-7b-instruct

Reviewed by RunLocalAI Editorial. See our editorial policy for how we research and verify model claims.

Related — keep moving

Before you buy

Verify E5 Mistral 7B Instruct runs on your specific hardware before committing money.