LFM2.5 24B A2B is a sparse mixture-of-experts model from LiquidAI that activates only 2 billion parameters per forward pass despite having 24 billion total parameters, making it notably efficient at inference time. It handles long contexts up to 131k tokens, which means it can work through lengthy documents without truncation. The trade-off is that sparse activation patterns can sometimes produce less consistent outputs than dense models of comparable active parameter counts.