Liquid AI releases open-weight LFM2.5-24B-A2B for on-device agents

Liquid AI released LFM2.5-24B-A2B on October 13, 2026, an open-weight 24B-parameter reasoning model with 2.45B active parameters for on-device agents. The context window grew from 32K to 131,072 tokens and the vocabulary doubled to 128K, with a 36-trillion-token training budget. It leads five of eight agentic benchmarks and decodes at 184 tokens/s on an Apple M5 Max.

• 24B total parameters, 2.45B active, 131,072-token context
• Trained on ~36T tokens, supports 16 languages
• 184 tokens/s on Apple M5 Max, 114 tokens/s on Ryzen AI Max+ 395
• API pricing: $0.10 per 1M input, $1.00 per 1M output tokens