text-embedding-ada-002 model availability in Azure OpenAI

Question

I'd like to know if and when the model text-embedding-ada-002 from OpenAI will be available in Azure OpenAI studio

Answer 1

Hello Marco Totolo

Thanks for reaching out to us, I have checked with product manager about the timeline.

The model will be available in February/before end of March. I am sorry I can not provide an exact date but only a range.

I hope this helps!

Regards,

Yutong

-Please kindly accept the answer if you feel helpful to support the community, thanks a lot.

Answer 2

Hi, it looks like this model is available now in East US which is great, however, it seems like it doesn't have the latest token size capabilities. According to Open AI...

Longer context. The context length of the new model is increased by a factor of four, from 2048 to 8192, making it more convenient to work with long documents.

...and I am getting errors...

2023-02-16 22:14:33,788 -INFO - openai - error_code=None error_message="This model's maximum context length is 2047 tokens, however you requested 3451 tokens (3451 in your prompt; 0 for the completion). Please reduce your prompt; or completion length." error_param=None error_type=invalid_request_error message='OpenAI API error received' stream_error=False

Why does the new model not match the Open AI specification? See https://openai.com/blog/new-and-improved-embedding-model/

Answer 3

Hi @YutongTie-MSFT

Azure OpenAI text-embedding-ada-002 still seem to support only 4k tokens while the original OpenAI model supports 8k.

Can you please find out when there will be parity in terms of token size?

Thank you!