text-embedding-3-large vs text-embedding-3-small
Same family, same API. The small model is close enough for English search and much cheaper; the large model is worth it when your documents are in more than one language, where its lead is wide.
Choose text-embedding-3-large if
- Documents in several languages — 54.9% on MIRACL against 44.0%.
- The highest retrieval accuracy OpenAI offers.
- Shorter vectors when you need to save storage.
Choose text-embedding-3-small if
- English search and recommendations.
- Very large volumes, where cost decides.
- An index already built for 1,536-number vectors.
Side by side
| text-embedding-3-large | text-embedding-3-small | |
|---|---|---|
| Made by | OpenAI | OpenAI |
| Released | January 2024 | January 2024 |
| Input price, per 1M tokens | ₹13.12 | ₹2.02 |
| Output price, per 1M tokens | — | — |
| Cached input, per 1M tokens | — | — |
| Batch (about half price) | Yes | Yes |
| Context window | 8K tokens | 8K tokens |
| Longest answer | — | — |
| Reads | Text | Text |
| Thinks before answering | No | No |
| Open weights | No | No |
Prices are Roar AI’s published rates, live from our price list. Specs are each lab’s own published figures. Prices are our published US dollar rates converted at ₹98.00 to the dollar, our rate for October — the same rate your card is charged.
Use both, switch any time
On Roar AI both are on the same key and the same monthly invoice, paid in your local currency. Trying the other one is a one-word change:
client.chat.completions.create(model="text-embedding-3-large", …) client.chat.completions.create(model="text-embedding-3-small", …)
Send the same prompts to both for a day and compare the answers on your own work — that settles it faster than any benchmark. Read the docs.
Try both
on one key
Send the same prompts to each and compare the answers on your own work.