# Ant Group's Inclusion AI Reports 5.8x to 6.9x Decoding Speedups From a Parallel Draft-Tree Method Called UBTree

The drafter builds trees of up to 64 candidate tokens that the target model checks in a single forward pass, and it beat the DARTree baseline in all 28 reported comparisons.

- Published: 2026-10-31T06:16:57.637Z
- Canonical: https://polylog.news/ai/2026-10-31/ant-group-s-inclusion-ai-reports-5-8x-to-6-9x-decoding-speed
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [Polylog editors](https://polylog.news), [arXiv](https://arxiv.org/abs/2609.39972), [GitHub (inclusionAI/ubtree)](https://github.com/inclusionAI/ubtree), [Hugging Face](https://huggingface.co/inclusionAI/Ling-3.0-tiny-UBTree), [GitHub (vLLM fork issue)](https://github.com/HanFa/vllm/issues/2)

Inclusion AI, the open-model group at Ant Group, has released UBTree, a speculative decoding method that builds the draft tree in parallel rather than one token at a time (AI ML Big Data). Standard tree drafters extend each branch autoregre…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-10-31/ant-group-s-inclusion-ai-reports-5-8x-to-6-9x-decoding-speed (subscription information: https://polylog.news/pricing).