Multilingual Open LLMs

A list of good open multilingual large language models.

Multilingual models are not as common as English ones. Note that this list is evolving and may not be up-to-date.

Name Description Primary languages Languages with likely a good support
Llama 3.1 Llama 3.1 405B is likely the best as of 2024-07-24. It is close to GPT-4o performance in English but not quite in other languages yet. EN, DE, FR, IT, PT, HI, ES, TH Likely a wide support. Speak more coherently in CS.
Mistral-Nemo-Instruct-2407 12B model trained for multilingualism. Seems to generate well in CS if prompted for. EN, CN, IT, FR, DE, ES, RU Likely a wide support. Perhaps 85% of all languages have some support.
Qwen2 Qwen2-72B and various other sizes Qwen2-0.5B, Qwen2-1.5B, Qwen2-7B, Qwen2-57B-A14B. EN, ZH EN, ZH, DE, FR, ES, PT, IT, NL, RU, CS, PL, AR, FA, HE, TR, JA, KO, VI, TH, ID, MS, LO, MY, CEB, KM, TL, HI, BN, UR
Gemma 2 It seems to be capable in Czech (CS), which a small language. I have seen similar reports for some other languages. Probably all more frequent languages than CS are supported. Based on the Whisper dataset these could be following EN ZH, DE, ES, RU, FR, PT, KO, JA, TR, PL, IT, SV, NL, CA, FI, ID, AR, UK, VI, HE, EL, DA, MS, HU, RO, NO, TH, CS.
Llama-3 Llama-3 is a multilingual model trained on over 100 languages. It is not intended for use with other languages than English. EN EN, FR, DE, ES, KO, AF, IT, CA, RO, CY, TJ, DA, VI.
Nemotron 4 Nemotron-4-15b, or Nemotron-4-340B-Instruct. I haven’t tested this model. EN ES, ID, NL, PL, FR, RU, IT, DE, VI, JA, TR, EL, RO, CS, AR, SV, FA, ZH, HU, KO, UK, NO, FL, HI, DA, BG, HR, SK, TH
EagleX 7B 2.25T Faster inference, but generally a bit lower quality. Not available on OpenRouter. I haven’t tested this model much. EN  

Created on 22 Jun 2024. Updated on: 18 Jul 2024.
Thank you










About Vaclav Kosar How many days left in this quarter? Twitter Bullet Points to Copy & Paste Averaging Stopwatch Privacy Policy
Copyright © Vaclav Kosar. All rights reserved. Not investment, financial, medical, or any other advice. No guarantee of information accuracy.