Google deprecates gemini-embedding-2-preview, routing requests to gemini-embedding-2
Google is retiring the gemini-embedding-2-preview model, and from November 6, 2026, its requests route to gemini-embedding-2 at the same rates.
Google is retiring the preview version of its Gemini embedding model. From November 6, 2026, any API request still sent to gemini-embedding-2-preview will be routed to gemini-embedding-2 and billed at that model's rates. The Gemini API changelog describes the change and says developers need only update their model string to gemini-embedding-2.
The practical effect depends on what the changelog says stays the same. According to the entry, gemini-embedding-2 uses the same API schema, the same pricing and the same embedding space as the preview model. The embedding space matters for anyone who has already stored vectors. Embeddings from two models are only comparable if they share a vector space, because similarity search compares distances between numbers that have no meaning outside their model. If the spaces differed, a developer would have to re-embed the existing corpus before queries could be matched against it. The changelog says no vector re-indexing is needed. Under its terms, the stored database keeps working and only the string sent with each request changes.
The entry does not say whether the routing is permanent or what happens after the preview name is removed entirely. It also gives no rate figures, so the billing point rests on the phrase "at Gemini Embedding 2 rates" rather than on a published price.
What changed
gemini-embedding-2-preview requests now go to gemini-embedding-2, billed at gemini-embedding-2 rates, and developers need only update their model string.
Type: Breaking change · Takes effect: Friday, November 6, 2026 · Applies to: gemini-embedding-2-preview, gemini-embedding-2
Why it matters
Teams whose retrieval pipelines call the preview embedding model will be served by the stable model automatically, and existing vectors stay compatible because the embedding space is unchanged.
What you can do
Update the model string from gemini-embedding-2-preview to gemini-embedding-2 in your code and configuration.
How the story developed
Google's own changelog confirms the deprecation and says the same schema, pricing and embedding space apply, so no re-indexing is needed.
Original source
Where we saw it
Gemini API changelog
AI wrote this summary from the sources listed, and nobody reviewed it individually before publication. Check the original source before you rely on it. How it works