OpenAI: Text Embedding Ada 002

OpenAI: Text Embedding Ada 002 is an embeddings model from OpenAI, listed on api14 under the model id openai/text-embedding-ada-002. It belongs to the embeddings category and is described by the provider as OpenAI's legacy text embedding model. If you have worked with OpenAI's embedding endpoints before, this is the familiar ada 002 generation of that line, made available through api14 alongside the rest of the catalogue.

The model takes text as input and returns embeddings as output, which means it does not generate prose or chat replies. It maps a piece of text to a vector representation that you can store and compare. Its context window is 8192 tokens, so a single request can cover a fairly long passage rather than a short snippet. The details list a maximum output of 7372, which reflects the shape of the vectors it returns rather than generated text.

Because the output is a vector, the practical uses are the standard embedding workflows. You can build semantic search over a document collection by embedding your corpus once and embedding each query at request time. You can use the vectors for clustering or deduplication, for example to group similar support tickets or find near-identical records. Classification on top of frozen embeddings is another common pattern: train a lightweight classifier on the vectors instead of fine-tuning a generative model. Retrieval augmented generation also fits here, where embeddings pick the relevant passages and a separate text generation model writes the answer.

Calling it through api14 works the same way as any other model on the platform, using the OpenAI-compatible API with openai/text-embedding-ada-002 as the model id. You send your text in the request and read the embedding back from the response. The supported parameters list is empty, so there are no extra tuning knobs to set beyond the basics of the call itself; you do not need to configure temperature, top_p or similar sampling options because this model does not sample text.

Keep the model's scope in mind when choosing it. It only accepts text, so images, audio and other modalities are out of range here. It produces embeddings only, so any task that needs an actual written answer requires a different model for the generation step. Since OpenAI describes it as a legacy embedding model, new projects may want to weigh it against newer embedding options, while existing pipelines already built around ada 002 vectors can continue to call it through api14 without changing their integration.