I haven’t tried it myself, but NLLB-200 has various sizes going down to 600M params: https://github.com/facebookresearch/fairseq/tree/nllb/ https://github.com/

I haven’t tried it myself, but NLLB-200 has various sizes going down to 600M params: https://github.com/facebookresearch/fairseq/tree/nllb/ https://github.com/facebookresearch/fairseq/tree/nllb/ If running locally is too difficult, you can use llm to access hosted models too.