Liquid AI released LFM2.5-24B-A2B on October 13, a mixture-of-experts reasoning model with 24 billion parameters in total and only 2 billion active per token.
Details
The model always reasons before answering, calls tools and handles multi-step tasks in popular agent harnesses. Architecture follows LFM2-24B-A2B (40 layers, 64 experts with top-4 routing), but the context window grows from 32K to 131K tokens and the vocabulary doubles to 128K tokens for better non-Latin text handling.
Liquid AI says it is competitive with Qwen3.6-35B-A3B on agentic tasks, scoring 48.09 on LiveClaw versus 38.19. It reports 184 tokens per second on an Apple M5 Max and up to 21K output tokens per second on a single NVIDIA B300 GPU. These are the company’s own figures.
Why it matters
With few active parameters, every reasoning token is cheap, which makes always-on local agents more practical.