[D] Any encoder only model having bigger max token than 512 (BERT, Roberta, etc)?

Thanks for reading this.

Is there any encoder model like BERT, Roberta, etc, thats has a bigger max token length, say around 4k. I am looking to finetune one for multi-class classification usecase.

Thanks in advance!

advance bert bigger class classification encoder etc machinelearning max reading roberta token