🤖LFM2.5-24B-A2B: 24B-Parameter Model for On-Device Agents
The biggest, fastest on-device reasoning model just dropped
TL;DR
LFM2.5-24B-A2B, a 24B-parameter model optimized for on-device agents, is now available on Hugging Face. It's the largest in the LFM2.5 family, running up to 1.7x faster than comparable models. Ideal for developers looking to enhance local AI capabilities.
LFM2.5-24B-A2B, a 24B-parameter reasoning model, is now the largest and final model in the LFM2.5 family, optimized for fast, private agents on-device. This model is a game changer for developers needing robust, on-device AI capabilities, with a 1.7x speed boost over similar models. It's available on Hugging Face and our Playground, with an expanded context window from 32,768 to 131,072 tokens and a vocabulary that doubles from 65K to 128K tokens. This makes it a must-have for anyone building local AI tools or agents.
-oOdUn61bDRSykk4H951SoDOHzqoOe1.png)
Key Points
LFM2.5-24B-A2B: 24B parameters, 2B active parameters, largest in LFM2.5 family
Scores 48.09 on LiveClaw, 38.19 on AutomationBench, 27.69 on ToolBench
Runs up to 1.7x faster than comparable MoEs on-device
Context window extended from 32,768 to 131,072 tokens
Vocabulary expanded from 65K to 128K tokens
Why It Matters
If you're building local AI agents or need fast, on-device reasoning, LFM2.5-24B-A2B is a game changer. With a 1.7x speed boost and expanded context window, it outperforms competitors on key benchmarks. Developers working with on-device AI will see significant improvements in performance and efficiency.
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.
Join 3,569 builders reading daily.