k
keynews.ai
keynews.ai
FrontierThe story, in brief

Liquid AI Releases LFM2.5-DSpark Draft Models That Deliver Up to 3.18x Faster Decoding Without Changing Model Outputs

3.18x faster decoding. Liquid AI's draft models show speculative decoding can scale without model changes.

Reporting by MarkTechPost 1 min brief
Illustration of a transparent lens revealing connected networks across layers of paper.
Exploring the next frontier of AI research.AI illustration by KeyNews
The KeyNews take

Why it matters

Speculative decoding—using smaller draft models to speed inference—is moving from research to production. Liquid AI's DSpark drafters maintain identical outputs while cutting latency, a key efficiency win for practitioners deploying LFM2.5 at scale.

The key facts

12 to know

  1. Liquid AI released LFM2.5-DSpark draft models

  2. Three ~300M parameter drafters

  3. Up to 3.18x faster decoding reported

  4. Identical greedy output—no model behavior changes

  5. Speculative decoding technique

  6. Published August 20, 2026

  7. 3.18x faster decoding measured

  8. Three ~300M parameter draft models released

  9. Identical greedy outputs (no quality loss)

  10. Technique: speculative decoding

  11. Model: LFM2.5 (Liquid Foundational Model 2.5)

  12. Published: August 20, 2026

Go to the source

MarkTechPostmarktechpost.com

Publisher excerpt: Three ~300M drafters bring speculative decoding to LFM2.5, delivering up to 3.18x faster decoding with identical greedy output.
Read original report
Back to today's editionMore frontier news

The wider picture

View all
Illustration of a transparent lens revealing connected networks across layers of paper.
AI illustration by KeyNews
Frontier01

Alibaba Unveils Zhenwu V900 — and Plans Qwen Models With Up to 10 Trillion Parameters

Alibaba is advancing on two fronts simultaneously: announcing a custom AI accelerator (Zhenwu V900) and committing to massive model scale (10T parameters for future Qwen releases). For practitioners, this matters as a credible third-party capability play outside the US-China licensing squeeze; for enthusiasts, it's a significant lab-race signal about training compute and parameter scaling as competitive levers.

TechRepublic
Illustration of a transparent lens revealing connected networks across layers of paper.
AI illustration by KeyNews
Frontier02

Kyutai Releases Voice of Reason: A Speech-Native Model that Solves Spoken Math with Reinforcement Learning

A new capability frontier: models that reason directly in speech without transcription bottlenecks. This changes how we think about multimodal reasoning and what's possible with open-weight releases at scale.

MarkTechPost
Illustration of a transparent lens revealing connected networks across layers of paper.
AI illustration by KeyNews
Frontier03

Claude Opus 5.5, GPT-6 Sol, GPT-6 Luna, and a new price war

Frontier labs are shipping upgraded reasoning and multimodal models in rapid succession, signaling acceleration in the capability race. Simultaneous price cuts reshape AI economics for practitioners.

Simon Willison