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

Sicherheitsmeldung mit Schweregrad noch nicht bewertet. Technische Details im Tab „Originaltext“; empfohlene Schritte in der Checkliste.

Originaltext der Quelle (nicht übersetzt) – die deutsche Zusammenfassung findest du im Tab „Kurzfassung".

Liquid AI has released DSpark draft model checkpoints for three models in its LFM2.5 family: LFM2.5-1.2B-Instruct, LFM2.5-2.6B, and LFM2.5-8B-A1B. Each drafter adds a speculative decoding path to an existing target model. A roughly 300M-parameter draft proposes a block of nine candidate tokens, and the target model verifies the whole block in a single forward pass. The trade is a small memory increase for a large decoding speedup: up to 3.18x on an H100 and up to 2.87x on an M4 Max MacBook Pro. Output does not change. Under greedy decoding, the emitted sequence is identical to the target model running alone, so benchmark accuracy is unchanged. Both llama.cpp and SGLang have day-one support.

Is it deployable?

Yes, if you self-host. The weights ship as Safetensors and GGUF, and the drafter checkpoints are not served by any hosted inference provider on Hugging Face today. Running them needs an SGLang or llama.cpp build with DSpark support for LFM2 targets.

Company level: The LFM Open License v1.0 allows free commercial use only while your entity stays under $10M in annual revenue. Indie developers, startups and SMBs are covered; larger enterprises must contact Liquid AI for a commercial license first.

Industries: Developer tooling, consumer apps that run locally, robotics and embedded systems, plus healthcare, finance and defense workloads that keep data on-premise or on-device.

Applications: Local coding assistants, on-device agents that reason before each tool call, single-user chat where batch size is 1, and offline copilots on laptop-class hardware.

What are …

Vollständiger Originalartikel ↗

• Betroffenheit im eigenen Stack prüfen: Versionen/Komponenten abgleichen.
• Originalquelle / Hersteller-Advisory lesen: https://www.marktechpost.com/2026/08/20/liquid-ai-releases-lfm2-5-dspark-draft-models-that-deliver-up-to-3-18x-faster-decoding/
• Verfügbaren Patch oder Workaround einspielen und dokumentieren.