feat(embedders)!: default to text-embedding-3-small

[camgrimsec]

Switches the default model for OpenAITextEmbedder, OpenAIDocumentEmbedder, AzureOpenAITextEmbedder, and AzureOpenAIDocumentEmbedder from text-embedding-ada-002 to text-embedding-3-small.

text-embedding-3-small is roughly 5x cheaper per token ($0.02 vs $0.10 per 1M tokens) and scores higher on MTEB. OpenAI marks ada-002 as legacy: https://platform.openai.com/docs/deprecations

Backward-compatible: users passing model= or azure_deployment= explicitly are unaffected. Release note added covering the embedding-incompatibility caveat and the Azure-deployment caveat.

Only default-assertion tests were updated; tests that explicitly construct embedders with ada-002 are preserved (it is still a valid model name).

Related Issues

• fixes #issue-number

Proposed Changes:

How did you test it?

Notes for the reviewer

Checklist

• I have read the contributors guidelines and the code of conduct.
• I have updated the related issue with new insights and changes.
• I have added unit tests and updated the docstrings.
• I've used one of the conventional commit types for my PR title: fix:, feat:, build:, chore:, ci:, docs:, style:, refactor:, perf:, test: and added ! in case the PR includes breaking changes.
• I have documented my code.
• I have added a release note file, following the contributors guidelines.
• I have run pre-commit hooks and fixed any issue.

[vercel]

@camgrimsec is attempting to deploy a commit to the deepset Team on Vercel.

A member of the Team first needs to authorize it.

[github-actions]

Coverage report

Click to see where and how coverage changed

File	Statements	Missing	Coverage	Coverage (new stmts)	Lines missing
haystack/components/embedders
openai_document_embedder.py
openai_text_embedder.py
Project Total

This report was generated by python-coverage-comment-action

[bogdankostic]

Thanks for the PR @camgrimsec! I left two comments.

Also, can you point me to the place in the OpenAI docs stating that text-embedding-ada-002 is deprecated?

[bogdankostic]

Let's keep this doc string short.

Suggested change

	The name of the model deployed on Azure. The default is `text-embedding-3-small`,
	which is roughly 5x cheaper than the legacy `text-embedding-ada-002` and scores higher
	on the MTEB benchmark. Note that this is a deployment name in your Azure resource,
	so the deployment must exist there. See OpenAI's
	[deprecation notice](https://platform.openai.com/docs/deprecations) for details.
	The name of the model deployed on Azure. The default is `text-embedding-3-small`.

[camgrimsec]

Applied

shortened to the one-line version you suggested

and made the same change in the other three embedders (OpenAITextEmbedder, OpenAIDocumentEmbedder, AzureOpenAITextEmbedder) for consistency.

[bogdankostic]

We need double backtick for in-line code in release notes.

[camgrimsec]

Fixed

All inline code in the release note now uses double backticks

and the deprecation-page reference was replaced with a link to OpenAI's new-embedding-models announcement

The original "legacy/deprecated" framing was inaccurate ada-002 is not on the formal deprecation schedule

[camgrimsec]

Thanks for pushing on the citation.

You're right.text-embedding-ada-002 is not on the deprecation schedule in OpenAI's docs. The accurate framing is that OpenAI introduced text-embedding-3-small as the recommended successor.

In their announcement (https://openai.com/index/new-embedding-models-and-api-updates/) they call ada-002 the "previous generation" model and state "We are not deprecating text-embedding-ada-002".

The motivation for the default switch is the 5x price reduction and higher MTEB score, not a deprecation timeline.

I've updated the PR to reflect that:

• Release note now links to the announcement instead of the deprecation page, and drops the "legacy" claim.
• Commit message reworded.
• Docstrings shortened per your inline suggestions (applied to all four embedders for consistency).
• Release note now uses double backticks for inline code.

Apologies