other
0.109B parameters
Commercial OK
Reviewed May 2026

all-mpnet-base-v2

all-mpnet-base-v2 is a 109M-parameter sentence-transformers embedder based on Microsoft's MPNet, producing 768-dim vectors with a 384-token context. Trained on the same 1B+ sentence pairs as all-MiniLM-L6-v2 but with a stronger backbone, it sits at the quality-tier above MiniLM-L6 and remains the most-cited 'general-purpose' sentence-transformers baseline.

License: apache-2.0·Context: 384 tokens
BLK · VERDICT

Our verdict

OP · Fredoline Eruo|VERIFIED MAY 29, 2026
unrated

The 'I want 768 dims, I want sentence-transformers, I want Apache-2.0' default. all-mpnet-base-v2 has been the reference English embedder since 2021 and remains a credible baseline today. For new 2026 projects, gte-modernbert-base or nomic-embed-text-v1.5 deliver better quality with longer context at lower parameter count — but the mpnet line is so embedded in production codebases that it earns a row regardless.

Overview

all-mpnet-base-v2 is a 109M-parameter sentence-transformers embedder based on Microsoft's MPNet, producing 768-dim vectors with a 384-token context. Trained on the same 1B+ sentence pairs as all-MiniLM-L6-v2 but with a stronger backbone, it sits at the quality-tier above MiniLM-L6 and remains the most-cited 'general-purpose' sentence-transformers baseline.

Strengths

  • MTEB English score ~57.8 — meaningfully above MiniLM-L6 at ~5x the params
  • 768-dim output matches the standard for vector DB schemas (Postgres pgvector default)
  • Apache-2.0, no usage restrictions
  • Mature ONNX/CoreML/safetensors/GGUF support across runtimes
  • Reference baseline in countless sentence-transformers tutorials and papers

Weaknesses

  • Loses to modern 100-150M embedders (gte-modernbert-base, nomic-v1.5) on MTEB by 5-8 points
  • 384-token context window is short by 2026 standards
  • English-only — no multilingual variant under the all-mpnet name
  • Symmetric similarity model; no query/document role distinction

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_M0.1 GB1 GB

Get the model

HuggingFace

Original weights

huggingface.co/sentence-transformers/all-mpnet-base-v2

Source repository — direct quantization required.

Hardware that runs this

Cards with enough VRAM for at least one quantization of all-mpnet-base-v2.

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.

Step down
Smaller — faster, runs on weaker hardware
No verdicted models in the next tier down yet.

Frequently asked

What's the minimum VRAM to run all-mpnet-base-v2?

1GB of VRAM is enough to run all-mpnet-base-v2 at the Q4_K_M quantization (file size 0.1 GB). Higher-quality quantizations need more.

Can I use all-mpnet-base-v2 commercially?

Yes — all-mpnet-base-v2 ships under the apache-2.0, which permits commercial use. Always read the license text before deployment.

What's the context length of all-mpnet-base-v2?

all-mpnet-base-v2 supports a context window of 384 tokens (about 0K).

Source: huggingface.co/sentence-transformers/all-mpnet-base-v2

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

Related — keep moving

Before you buy

Verify all-mpnet-base-v2 runs on your specific hardware before committing money.