e5-base-v2

Mid-sized E5 embedding checkpoint for English retrieval and semantic search with a more practical serving footprint than the large variant.

Microsoft / intfloat · E5

Editorial review

Reviewed byOpenSourcesAI EditorialLast updatedAugust 2026SourcesExact model card

Model checkpoints, context windows, provider support, local runtime compatibility, and license terms can change quickly. Verify the exact model card before production or commercial use.

Best for

Builders who want a balanced English embedding model for search and RAG on moderate local or hosted infrastructure.

Who should use it

Builders who want a balanced English embedding model for search and RAG on moderate local or hosted infrastructure.
Builders who want local or self-hosted testing options.
RAG builders comparing retrieval quality, latency, and multilingual coverage.

Common workflows

Balanced embedding workflows
embedding workflows
rag workflows
semantic search workflows
retrieval workflows

Deployment and hardware notes

~0.44 GB in fp32 (110M parameters) — comfortable CPU serving for moderate workloads; a GPU mostly accelerates batch indexing.

License and usage notes

MIT. Open weights. Verify the exact model card and license terms for the checkpoint or hosted provider you use.

Strengths

Open weights model option for E5 workflows.
Builders who want a balanced English embedding model for search and RAG on moderate local or hosted infrastructure.
'query: ' on search strings, 'passage: ' on documents — the prefixes are part of the model's training contract, not optional formatting.

Limitations

English-only retrieval with 512-token inputs, and quality depends on remembering the E5 prefixes on every string you embed.
~0.44 GB in fp32 (110M parameters) — comfortable CPU serving for moderate workloads; a GPU mostly accelerates batch indexing.
Context window and limits: 512 tokens.
Verify the exact model card, provider docs, license, and serving support before production use.

Local workflow notes

'query: ' on search strings, 'passage: ' on documents — the prefixes are part of the model's training contract, not optional formatting.

Local runtimes: Ollama where supported, LM Studio where supported, llama.cpp where supported, Transformers

Platforms: Windows, macOS, Linux

Embedding spec

Memory~0.44 GB in fp32 (110M parameters) · ~0.22 GB in fp16Output768-dimension vectorsMax input512 tokens

The middle English v2 E5 at 768 dimensions: a practical default when e5-small-v2's quality falls short but e5-large-v2's index size and latency are unwelcome.