[deskamess]

Do they have a good multilingual embedding model? Ideally, with a decent context size like 16/32K. I think Qwen has one at 32K. Even the Gemma contexts are pretty small (8K).