What Speculators v0.5.0 adds for vLLM speculative decoding: DFlash block-diffusion draft models, unified online and offline training, native hidden-state extraction, and Gemma 4 latency results.
Speculators v0.5.0: DFlash Support and Online Training
calendar_today
May 28, 2026
person
Fynn Schmitt-Ulms, Helen Zhao, Rahul Tuli and Dipika Sikka (Red Hat AI Model Optimization Team)
domain
vllm