3 ms·
I wouldn’t say most—maybe a factor of 2. Getting the embedding is still an API call to an LLM.
by mdagostino 4y ago
I wouldn’t say most—maybe a factor of 2. Getting the embedding is still an API call to an LLM.
- DanielVZ 4y agoI’m pretty sure they were using a high cost LLM to summarize, and for embeddings you only need Ada, which is orders pf magnitude cheaper.