LLM Digest

Story

vllm_blog · Jul 13, 2026 · news

Source brief

EAGLE-3 Speculative Decoding on AMD Instinct GPUs: Training and Serving with vLLM and AMD Quark

vllm.aiJul 13, 2026
original source linked

In brief

How AMD Quark trains, quantizes, and serves EAGLE-3 speculative-decoding drafts with vLLM on AMD Instinct GPUs, delivering up to 2.00x throughput gains for Kimi-K2.5 and 1.79x for MiniMax-M2.5.

Read the original at vllm.ai →Open in live feed Read that day’s brief

EAGLE-3 Speculative Decoding on AMD Instinct GPUs: Training and Serving with vLLM and AMD Quark

Earlier in this thread 4 items

Achieve state-of-the-art inference latencies with speculative decoding

Boost Inference Performance up to 15x on NVIDIA Blackwell Using DFlash Speculative Decoding - NVIDIA Developer

[AINews] GLM-5.2: the top Frontend Coding model in the world, IndexShare for Speculative Decoding

Meet EAGLE 3.1: The Speculative Decoding Algorithm That Fixes Attention Drift in LLM Inference - MarkTechPost