Liquid AI released DSpark draft model checkpoints for three LFM2.5 models, enabling speculative decoding to speed up inference
Model release Provisional 86% confidence first seen
Liquid AI published DSpark draft model checkpoints for three LFM2.5 models that add a speculative-decoding path intended to accelerate inference while preserving greedy output quality. The coverage reports substantial latency and throughput improvements on hardware benchmarks, and notes day-one DSpark support in llama.cpp and SGLang for using the checkpoints.