qwen3-embed-8b — open-weights model published with reproducible eval cards.
llama-instruct — open-weights model published with reproducible eval cards.