50-language pre-trained base
According to the model card, this is the X-ALMA 13B multilingual pre-trained base model covering 50 listed languages.
Open Source Model Profile · haoranxu
X-ALMA-13B-Pretrain
X-ALMA-13B-Pretrain is a 13.02B-parameter Llama multilingual base model from haoranxu. Its model card documents 50-language pre-training and plug-and-play translation modules.
- Publisher
- haoranxu
- Task
- text-generation
- Model type
- llama
- License
- mit
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
X-ALMA-13B-Pretrain is published by haoranxu as a Llama-based multilingual text-generation base model. The captured configuration identifies LlamaForCausalLM and Safetensors metadata reports 13,015,864,320 parameters. According to the model card, it expands ALMA-R from 6 to 50 languages with plug-and-play modules, and this checkpoint is the 13B pre-trained base.
Recorded capabilities
Plug-and-play module design
According to the model card, X-ALMA uses language-specific modules with eight group checkpoints that can be merged or loaded separately.
Documented translation flow
According to the model card, translation adds the source sentence to a language-direction prompt and generates with beam search and sampling controls.
Three loading patterns
According to the model card, users can load a merged model, a base-plus-module PEFT setup, or the full multi-module assembly.
Broad language tagging
Hub tags list dozens of language codes alongside transformers, safetensors, and endpoints-compatible text generation.
Use cases in the source record
- Translation workflows that place a source sentence in the documented language-direction prompt and generate target text.
- Multilingual open-ended QA experiments that the card says the model should also support.
Limitations and unknowns
- No evaluation results were extracted from this record.
- No context-window value was extracted from this record.
- Provider state is historical snapshot data, not independently refreshed current availability.
- Language coverage and module behavior come from the publisher model card and have not been independently verified by Ethen.
Source and provenance
Source: haoranxu/X-ALMA-13B-Pretrain
Captured: Unknown. Processed: 2026-09-07T19:34:46.455508+00:00.
X-ALMA builds upon ALMA-R by expanding support from 6 to 50 languages. It utilizes a plug-and-play architecture with language-specific modules, complemented by a carefully designed training recipe. This release includes the X-ALMA pre-trained base model . @misc{xu2024xalmaplugplay, title={X-ALMA: Plug & Play Modules and Adaptive Rejection for Quality Translation at Scale}, author={Haoran Xu and Kenton Murray and Philipp Koehn and Hieu Hoang and Akiko Eriguchi and Huda Khayrallah}, year={2024}, eprint={2410.03115}, archivePrefix={arXiv}, primaryClass={cs.CL}, url={https://arxiv.org/abs/2410.03115}, } X-ALMA-13B-Pretrain is pre-traine…
F001F002F003F004F005F006F007F008F009F010F011F012F013F014F016F019