inclusionAI releases Ling3-Tiny-UBTree speculative-decoding draft model for Ling-3.0-tiny
AIinclusionAI has released Ling3-Tiny-UBTree-B16 on Hugging Face, a 183.1M-parameter draft model that proposes multiple candidate continuations for Ling-3.0-tiny to verify during speculative decoding. The checkpoint uses a five-layer DFlash backbone with a learned bigram selector, stores BF16 weights, and cannot generate text on its own.
Source details
inclusionAI (Ant Ling) · new models on Hugging Face · huggingface.coPublished · added here
Original title: inclusionAI/Ling-3.0-tiny-UBTree