# How Lossless Is Lossless Speculative Decoding? The Role of Numerical Precision in Orthrus

> Source: <https://aiflash.com/news/119973/>
> Published: 2026-09-15 11:30:00+00:00

Orthrus is a hybrid autoregressive-diffusion architecture that accelerates autoregressive language-model inference by generating multiple tokens in parallel while using a frozen autoregressive backbone. Its central claim is that an intra-model consensus mechanism enables lossless speculative decodin