other
0.137B parameters
Commercial OK
Reviewed May 2026

Nomic Embed Text v1.5

Nomic Embed Text v1.5 is a 137M-parameter English embedding model with an 8192-token context window, trained with Matryoshka Representation Learning so the 768-dim output can be truncated to 64/128/256/512 dims with minimal quality loss. The training data, code, and weights are fully open under Apache 2.0, and the model outperformed text-embedding-ada-002 and text-embedding-3-small on long-context MTEB at release.

License: apache-2.0·Context: 8,192 tokens
BLK · VERDICT

Our verdict

OP · Fredoline Eruo|VERIFIED MAY 29, 2026
unrated

The default open-weight embedding pick for English RAG in 2026. Matryoshka truncation is genuinely useful — you can ship 256-dim vectors to Postgres and reindex at 768 only for cold tiers. Fully open license and training recipe make it the safest production choice when provenance matters. Use jina-v3 instead if you need multilingual.

Overview

Nomic Embed Text v1.5 is a 137M-parameter English embedding model with an 8192-token context window, trained with Matryoshka Representation Learning so the 768-dim output can be truncated to 64/128/256/512 dims with minimal quality loss. The training data, code, and weights are fully open under Apache 2.0, and the model outperformed text-embedding-ada-002 and text-embedding-3-small on long-context MTEB at release.

Strengths

  • 768-dim Matryoshka output truncatable to 64/128/256/512 dims for storage/cost tradeoffs
  • Native 8192-token context via Rotary + dynamic NTK scaling — rare among sub-200M embedders
  • Fully reproducible: training data, code, and weights all Apache-2.0 open
  • Beats OpenAI text-embedding-ada-002 on MTEB and LoCo long-context benchmarks

Weaknesses

  • English-only; multilingual queries should use jina-v3 or arctic-l-v2 instead
  • Requires task-prefix discipline (search_query: / search_document:) — silent quality loss without it
  • Custom nomic_bert architecture needs trust_remote_code=True in transformers
  • MTEB score (~62.3) trails 7B-class embedders by ~4 points

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/nomic-ai/nomic-embed-text-v1.5

Source repository — direct quantization required.

Hardware that runs this

Cards with enough VRAM for at least one quantization of Nomic Embed Text v1.5.

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 up
More capable — bigger memory footprint
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 Nomic Embed Text v1.5?

1GB of VRAM is enough to run Nomic Embed Text v1.5 at the Q4_K_M quantization (file size 0.1 GB). Higher-quality quantizations need more.

Can I use Nomic Embed Text v1.5 commercially?

Yes — Nomic Embed Text v1.5 ships under the apache-2.0, which permits commercial use. Always read the license text before deployment.

What's the context length of Nomic Embed Text v1.5?

Nomic Embed Text v1.5 supports a context window of 8,192 tokens (about 8K).

Source: huggingface.co/nomic-ai/nomic-embed-text-v1.5

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

Related — keep moving

Before you buy

Verify Nomic Embed Text v1.5 runs on your specific hardware before committing money.