---
type: "raw_signal"
id: "basket-713652c2-29"
slug: "liquid-ai-lfm2-5-dspark-runtime"
title: "Liquid AI Uses DSpark to Accelerate LFM2.5 Decoding"
description: "Liquid AI has released DSpark draft checkpoints for three LFM2.5 models to accelerate decoding without changing the target model's output distribution."
retrieval_nugget: "Liquid AI has released DSpark draft checkpoints for three LFM2.5 models to accelerate decoding without changing the target model's output distribution. The systems lesson is hardware-aware verification rather than a free model upgrade. Acceptance rate, batch capacity, memory overhead, and target workload determine the realized gain. Teams should reproduce throughput and quality parity on their own prompts and serving configuration before adopting the headline multiplier."
observed_at: "2026-08-27"
record_date: "2026-08-27"
date_kind: "observed_at"
why_it_matters: "The systems lesson is hardware-aware verification rather than a free model upgrade. Acceptance rate, batch capacity, memory overhead, and target workload determine the realized gain. Teams should reproduce throughput and quality parity on their own prompts and serving configuration before adopting the headline multiplier."
novelty: "notable"
verification_level: "source-linked"
signal_type: "release"
source_platform: "liquid.ai"
topics: ["new-models","new-feature"]
entities: []
related_patterns: ["harness-architecture-outlives-model-choice"]
source_url: "https://liquid.ai/blog/lfm2.5-dspark"
source_urls: ["https://liquid.ai/blog/lfm2.5-dspark"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "signal:liquid-ai-lfm2-5-dspark-runtime"
status: "published"
visuals: []
routes: {"html":"https://newruntime.com/signals/liquid-ai-lfm2-5-dspark-runtime/","markdown":"https://newruntime.com/signals/liquid-ai-lfm2-5-dspark-runtime.md","json":"https://newruntime.com/signals/liquid-ai-lfm2-5-dspark-runtime.json"}
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
---

# Liquid AI Uses DSpark to Accelerate LFM2.5 Decoding

## Retrieval answer

Liquid AI has released DSpark draft checkpoints for three LFM2.5 models to accelerate decoding without changing the target model's output distribution. The systems lesson is hardware-aware verification rather than a free model upgrade. Acceptance rate, batch capacity, memory overhead, and target workload determine the realized gain. Teams should reproduce throughput and quality parity on their own prompts and serving configuration before adopting the headline multiplier.

Liquid AI has released DSpark draft checkpoints for three LFM2.5 models to accelerate decoding without changing the target model's output distribution.

The speculative path uses a lightweight draft model to propose tokens and lets the target verify a block in one forward pass. Liquid reports throughput gains up to 3.18 times on GPU and 2.87 times on-device, with integrations released for llama.cpp and SGLang.

The systems lesson is hardware-aware verification rather than a free model upgrade. Acceptance rate, batch capacity, memory overhead, and target workload determine the realized gain. Teams should reproduce throughput and quality parity on their own prompts and serving configuration before adopting the headline multiplier.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [New Models](https://newruntime.com/topics/new-models.md): Explore the new-models topic hub.
- topic hub: [New Feature](https://newruntime.com/topics/new-feature.md): Explore the new-feature topic hub.
- collection: [Raw Signals](https://newruntime.com/signals.json): Open every source-linked raw signal.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)