PortBERT: Navigating the Depths of Portuguese Language Models
2026-06-01 • Computation and Language
Computation and Language
AI summaryⓘ
The authors created PortBERT, a set of Portuguese language models based on RoBERTa, to balance good performance with efficient training and usage. They trained these models from scratch on a very large, cleaned dataset and tested them on Portuguese versions of standard language tasks, where they did as well or better than other models. Apart from accuracy, the authors also shared details about how long training and using the models take, helping others understand their efficiency. They made their models freely available for researchers and developers to use.
Transformer modelsRoBERTaPortuguese NLPByte-level BPE tokenizationmC4 datasetOSCAR datasetExtraGLUE benchmarkModel efficiencyfairseqHuggingface
Authors
Raphael Scheible-Schmitt, Henry He, Armando B. Mendes
Abstract
Transformer models dominate modern NLP, but efficient, language-specific models remain scarce. In Portuguese, most focus on scale or accuracy, often neglecting training and deployment efficiency. In the present work, we introduce PortBERT, a family of RoBERTa-based language models for Portuguese, designed to balance performance and efficiency. Trained from scratch on over 450 GB of deduplicated and filtered mC4 and OSCAR23 from CulturaX using fairseq, PortBERT leverages byte-level BPE tokenization and stable pre-training routines across both GPU and TPU processors. We release two variants, PortBERT base and PortBERT large, and evaluate them on ExtraGLUE, a suite of translated GLUE and SuperGLUE tasks. Both models perform competitively, matching or surpassing existing monolingual and multilingual models. Beyond accuracy, we report training and inference times as well as fine-tuning throughput, providing practical insights into model efficiency. PortBERT thus complements prior work by addressing the underexplored dimension of compute-performance tradeoffs in Portuguese NLP. We release all models on Huggingface and provide fairseq checkpoints to support further research and applications.