Key Info
Ollama has added LFM2.5-24B-A2B to its library, the largest model in Liquid AI's LFM2.5 family, positioned as an efficient local model for running multi-step agent tasks on consumer laptops and PCs.
Highlights
- 24B total parameters with only 2B active per token, keeping inference efficient on local hardware
- Context window expanded to 131K tokens for longer agent trajectories
- 128K-token vocabulary for improved tokenization efficiency
- Designed to work inside popular agent harnesses for multi-step task execution
- Available via
ollama run lfm2.5:24b