← All model drops
Live drop
LFM2.5-24B-A2B
Liquid AIDropped Oct 13, 2026
24B-total / ~2B-active MoE reasoning model optimized for fast, private on-device agents and tool use.
About this release
Live drop
Liquid AIDropped Oct 13, 2026
24B-total / ~2B-active MoE reasoning model optimized for fast, private on-device agents and tool use.
About this release
LFM2.5-24B-A2B is the largest model in Liquid AI's LFM2.5 family: a sparse MoE with 24B total parameters and ~2B active per token, 131K context, and explicit reasoning before answers. It matches or leads larger models on multiple agentic/tool-use benchmarks while running efficiently on consumer hardware like Apple M5 Max at up to 184 tokens/s. Open weights (base and post-trained) are available on Hugging Face under the LFM license.
Read the official Liquid AI launch pageFirst builds incoming — this lane fills in automatically as people start building with LFM2.5-24B-A2B. Know one? Submit the URL.