OpenAI GPT-6 model for focused, high-volume workloads where cost efficiency is the priority.
text-embedding-3-small
Highly efficient embedding model featuring Matryoshka dimension shortening for ultra-low storage and compute costs.
Technical Specifications
Benchmark Evaluations
Deep Architectural Overview
text-embedding-3-small generates 1536-dimensional embeddings for vector search and RAG. Using Matryoshka Representation Learning, developers can shorten vectors to 512 dimensions without substantial loss of retrieval accuracy.
Strengths & Considerations
- Ultra-cheap $0.02 per 1M tokens
- Matryoshka representation shortening down to 512 dims
- Outperforms text-embedding-ada-002
- Complex multi-lingual search is better served by text-embedding-3-large
Token & API Pricing
Similar & Alternative Models
Explore other frontier models from OpenAI and comparable reasoning engines.
OpenAI GPT-6 model built for complex coding and agentic workflows, balancing intelligence and cost.
OpenAI’s premier sixth-generation frontier model, built for the most demanding end-to-end software engineering and scientific research tasks.
Flagship model family of mid-2026 featuring Sol (flagship), Terra (balanced), Luna (high-speed), and Cyber (authorized vulnerability research).