text-embedding-3-large
Embedding modelCard summarizedMost capable embedding model for both english and non-english tasks
Source: "New embedding models and API updates" launch post and Embeddings guide (no model card published)
Where to call it
- OpenAI API
text-embedding-3-large
Good for
- Search (ranking results by relevance to a query)
- Clustering and classification of text
- Recommendations and anomaly detection
Overview
OpenAI · 02Large, front
Embed 3 Large
Most capable embedding model for both english and non-english tasks
Typical price
$0.13 / 1M tok
Text input
- Dimensions
- 3072
- Max input
- 8,192 tokens
- Modes
- text embeddings
- Released
- 25 Jan 2024
text-embedding-3-large is OpenAI's larger third-generation text embedding model, returning vectors of up to 3072 dimensions. OpenAI calls it its best-performing embedding model and reports large gains over text-embedding-ada-002 on multilingual retrieval (MIRACL) and English tasks (MTEB). Both third-generation models were trained so the vector can be shortened with the dimensions parameter while keeping its meaning. launch post; embeddings guide
text-embedding-3-large- Typical price
- $0.13 / 1M tok
- Dimensions
- 3072 (can be shortened with the dimensions parameter)
- Max input
- 8,192 tokens
- Languages
- Multilingual (MIRACL multi-language retrieval reported); no language count stated
- Weights
- Closed
- Released
- 25 Jan 2024
- text embeddings