text-embedding-ada-002
by OpenAI
Announced
2024-01-25
Complete
—
Refund window closes
—
Refund status
No refund applicable
OpenAI's text-embedding-ada-002 was the de-facto standard embedding model from late 2022 through 2023, powering most RAG (retrieval-augmented generation) applications. In January 2024, OpenAI released the text-embedding-3 family (small and large variants) with significantly better quality, lower cost, and configurable output dimensions. Ada-002 was placed in deprecated status. The endpoint still responds but receives no further improvements, and apps using ada-002 will benefit from migration.
Refund flow
- 1
Not applicable. OpenAI API is pay-as-you-go — no pre-paid credits tied to embedding models.
Migration path
Recommended successor
text-embedding-3-small or text-embedding-3-large
Why this is the closest fit
Direct successor models from the same family. -3-small is faster + cheaper than ada-002; -3-large is better quality.
What differs from the original
Different embedding dimensions (1536 default for ada-002 vs 1536/3072 for -3 family). Migration requires re-embedding your entire vector store — embeddings are not cross-model compatible. Output quality on retrieval tasks is meaningfully better with -3.
What this tool meant
text-embedding-ada-002 was the embedding model that made consumer-tier RAG economically viable. Released in late 2022 with a 100x cost reduction over the prior `text-search-davinci-001`, it triggered the wave of "chat with your documents" SaaS products in 2023. Its successor (text-embedding-3) is meaningfully better but the ecosystem cost to re-embed every existing vector store was substantial — many apps still run ada-002 in production well into 2025.
Sources
Spent credits on AI video that failed?
AVA automates the refund flow across every video provider
Free Chrome extension. Detects the failure mode, captures evidence, drafts the refund email with the technical term and Generation ID. Click send.