Recommended models

[shssoichiro]

After looking through the latest release notes and seeing the changes to the list of recommended models, I'm a bit perplexed by the lack of inclusion of open models for the embed and reranker role. nomic-embed-text is widely considered a strong embedding model, and there are also qwen3-embedding and qwen3-reranker available at 8b, 4b, and 0.6b sizes, with qwen3-embedding even at 0.6b performing competitively with nomic-embed-text but providing much larger context windows. It should by no means be required to use a closed-source embedding or reranking model in my opinion.

[bdougie]

These seem reasonable and we'd be happy to take a pr to include these. The list is currently incomplete but was needed to support some open questions from users

[shssoichiro]

Thanks! I can create a PR to add those then, I just wasn't sure if there was a reason those were being excluded.

One hiccup may however be that there are no official configs on the hub for qwen3-embedding and reranker. Would the Ollama org be open to looking into these? (I'm not sure how their processes go or if your team has any collaboration with theirs, or if they add models to the hub fully independently.)

[bdougie]

I don't think it is necessity to link to the Hub in all cases. List the model in the column is all that is required. A link to the hub is a nice to have if it exist. The ollama can update the hub with these listings, but I think it would be more useful after we update the listings.

[shssoichiro]

Closing this discussion as #7313 has been opened to resolve it.