LFM2.5 (24B-A2B, Q4)

Liquid AI's largest LFM2.5, a mixture-of-experts model that holds 24B parameters but uses only about 2B for each word it writes, so it stays quick for its size. Built for agent work: it always thinks before it answers, calls tools natively, and has a 128K context window. Pair it with search or your documents for fact-heavy questions.

Gezel downloads this model the first time a gezel needs it, checks the file against its published hash, and applies tuned settings for it.

The model’s own page · Every model · Model scorecard