---
title: "text-embedding-3-small Notes"
url: https://memory.wiki/ycd06RQQ
updated: 2026-05-10T21:00:23.747Z
hub: https://memory.wiki/hub/yc-demo
concept_count: 12
source: "memory.wiki"
---
# text-embedding-3-small Notes

OpenAI's most cost-effective embedding model.

- **Dimensions**: 1536 (default), reducible via `dimensions` param
- **Pricing**: $0.02 / 1M tokens input
- **Context**: 8191 tokens
- **Use case**: semantic search, clustering, recommendation

## Trade-offs vs -3-large
- Smaller = cheaper + faster
- Quality drop is real but small for general English
- For multilingual: -large pulls ahead

## Why mdfy uses -small
50,000 docs × ~1k tokens/doc × $0.02/M = $1 to fully embed a hub.
That's the right operating point for free-tier users.

## Reduction trick
Truncate to 512 dims for storage savings — re-normalize before
similarity search. mdfy doesn't do this yet; pgvector(1536) is fine
under 100k docs/user.


---

## Concepts in this document
- **mdfy** _(entity)_
  Alternative approach emphasizing author-owned memory and cross-platform sharing via URLs.
- **pgvector** _(entity)_
  Vector storage technology used by mdfy for managing 1536-dimensional embeddings.
- **semantic search** _(tag)_
  Primary application use case for text-embedding-3-small in production systems.
- **text-embedding-3-small** _(entity)_
  OpenAI's cost-effective embedding model that is the primary subject and recommended choice for the mdfy platform.
- **embedding model** _(concept)_
  The core technology being evaluated and documented for semantic representation of text.
- **text-embedding-3-large** _(entity)_
  Competing model explicitly compared on quality and multilingual performance tradeoffs.
- **cost-effectiveness** _(concept)_
  Central optimization criterion driving mdfy's choice of this model at $0.02 per 1M tokens.
- **cost-efficiency** _(concept)_
  Central decision criterion: $0.02 per 1M tokens pricing makes -small viable for free-tier users at scale.
- **multilingual support** _(concept)_
  Known weakness of text-embedding-3-small where -large performs noticeably better.
- **dimension reduction** _(concept)_
  Optimization technique to truncate embeddings from 1536 to 512 dimensions for storage efficiency.
- **quality-speed tradeoff** _(concept)_
  Design tension between model capability and inference performance that justifies choosing the smaller variant.
- **vector dimensions** _(concept)_
  Technical configuration parameter affecting storage and performance; can be reduced from 1536 default.

## Concept relations (within this doc's concepts)
- **semantic search** implemented via **pgvector**
- **text-embedding-3-small** optimizes for **cost-effectiveness**
- **text-embedding-3-small** enables **semantic search**
- **mdfy** adopts **text-embedding-3-small**
- **text-embedding-3-small** compared to **text-embedding-3-large**
- **dimension reduction** improves **cost-effectiveness**
- **text-embedding-3-small** demonstrates **quality-speed tradeoff**
- **text-embedding-3-large** excels at **multilingual support**
- **mdfy** stores embeddings in **pgvector**
- **text-embedding-3-small** trades quality for cost **text-embedding-3-large**
- **text-embedding-3-small** enables via pricing **cost-efficiency**
- **text-embedding-3-small** performs worse at **multilingual support**
- **text-embedding-3-small** primary use case **semantic search**
- **mdfy** selected for adoption **text-embedding-3-small**

_Hub canonical:_ https://memory.wiki/hub/yc-demo
_Concept digest:_ https://memory.wiki/raw/hub/yc-demo?digest=1&compact=1