Liquid AI Launches LFM2.5-2.6B On-Device Agent Model

alexcovo_eth · x · 2026-08-05

Liquid AI has released LFM2.5-2.6B, a 2.6 billion parameter dense hybrid model. Designed specifically as an on-device agent, it runs locally to ensure data privacy and marginal costs of roughly zero.

Pretrained on 34T tokens, the model supports a 128K context length. It offers day-one compatibility with popular inference frameworks like llama.cpp, MLX, and vLLM. It has been post-trained inside real agent harnesses, equipping it with multi-step planning and tool-using capabilities.

Related event: Liquid AI Releases 2.6B On-Device Agent Model LFM2.5(18 posts)→

Original post →

More from Models

Models channel →