Back to AI News
arxiv

Lara Mini Base: A Multilingual MoE Base Model Trained on 6T Tokens

Lara Mini Base is a 7.2B-parameter mixture-of-experts model with 0.6B active parameters, trained from scratch on 6T tokens spanning English and 44 other languages, covering all 24 official EU languages. It supports 128k-token contexts and is released under the CC BY-NC-SA 4.0 license. The authors report it achieves the highest average score among open base models with up to 2B active parameters on English, multilingual, math, and code benchmarks, with the largest margins on multilingual and math tasks. They also plan to release T-HellaSwag, a machine-translated commonsense reasoning benchmark in 110 language variants.

AI Brief

This is a concrete new model release with a novel multilingual MoE design and benchmark claims relevant to efficient multilingual AI.

arXiv:2611.16956v1 Announce Type: cross Abstract: General-purpose language models are trained mostly on English: over open models we review, at most, non-English pretraining data makes up only a 12% share. Models built for European languages give them far more, but score well below general-purpose models of the same size. A common concern is that, in small models, languages compete for capacity. We present Lara Mini base, a model trained from scratch on 6T tokens in English and 44 other languages, which together cover all 24 official languages of the European Union. As training progresses, the 44 other languages together come to outweigh English. To give these languages enough capacity at low cost, Lara Mini base uses a mixture of experts architecture with 7.2B total and 0.6B active parameters. Among open base models with up to 2B active parameters, it achieves the highest average score across standard English, multilingual, math, and code benchmarks, ahead of models trained on 4 to 6 times as many tokens. Its margin is largest on multilingual benchmarks in the languages it supports and on math benchmarks. It supports contexts of up to 128k tokens. We find that our MoE architecture reduces capacity competition between languages. For a single user on one GPU, without quantization, it generates text at least 1.8 times faster than any model with a higher average score. We release the weights of Lara Mini base under the CC BY-NC-SA 4.0 license. In addition, we will release T-HellaSwag, a higher-quality machine translation of the commonsense reasoning benchmark HellaSwag into 110 language variants.
Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!