Launches
Liquid AI Drops a 24B Local Agent Model With Only 2B Active Parameters
Developing 2 sources
L
LFM2.5-24B-A2B
Model Oct 13, 2026
Liquid AI released a 24B-parameter, 2B-active open-weight reasoning model for local agent and tool-use workflows.
Visit primary link
2 sources attached
What happened lately
LFM2.5-24B-A2B timeline
Liquid AI released LFM2.5-24B-A2B on October 13 with downloadable base and instruction-tuned weights. The mixture-of-experts model has 24 billion total parameters and activates about 2 billion per token. Liquid AI positions it for local agents that reason and call tools, with support documented for several inference engines. Its model card lists the custom LFM1.0 license rather than a standard permissive open-source license. Reported speed and benchmark results are Liquid AI’s measurements; the card also advises using retrieval for knowledge-heavy tasks.