Model Usage Ada $0.0004 / 1K tokens
Roughly speaking, a company with 1,000 employees posting an average of 1,000 words daily to the groupware would pay less than $200 per day for embedding every utterance.
Assuming an average of 10 posts, that’s 60MB of vector indexes created per day, so Pinecore’s pay-as-you-go fee is 1,000 in running costs.
New and improved embedding model
2022-12-15 text-embedding-ada-002 is best
- context length: 8192
- 1536 dimensions
- They made it 1/8th the size, but they still got more talent.
OpenAI API:what-are-embeddings
- cl100k_base is the name of the tokenizer
-
For second-generation embedding models like text-embedding-ada-002, use the cl100k_base encoding.
- tiktoken
-
- Pinecore
This page is auto-translated from [/nishio/embeddingは0.4 / 1M Token) using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I’m very happy to spread my thought to non-Japanese readers.