A faithful safetensors repackaging of the pristine upstream facebook/nllb-200-distilled-1.3B — same fp32 weights (vocab 256206, d_model 1024, 24+24 layers), config, and tokenizer, converted to sharded safetensors. The 1.3B counterpart to MonumentalSystems/nllb-200-3.3B.

No fine-tuning or vocabulary changes; identical outputs to the upstream model. Licensed CC-BY-NC-4.0 per the upstream NLLB-200 release.