News/Models
Try nowModels·IncrementalOfficialstableUpdated Oct 13·Updated 2×·Event Oct 10, 2026·First seen Oct 13
Liquid AI Releases lfm2.5-24b-a2b Model for On-Device Agents
LaunchOpen sourcePractical
Try it
Something you can actually use or run today.
On October 10, 2026, Liquid AI released the LFM2.5-24B-A2B model designed for local, private execution of agentic tasks.
It provides a middle-ground model that is small enough for modern workstations but large enough to handle tool-use reliably.
AILookup take
Liquid AI is carving out a niche in 'privacy-first' agentics. This is a solid tool for developers who cannot send data to third-party APIs but need more reasoning than a 7B model provides.
Who cares
privacy-focused developerslocal LLM usersedge computing builders
Watch next
Integration of this model into popular local frameworks like Ollama or LM Studio.
Details
- Enables sophisticated agentic workflows like reasoning and tool use locally without cloud dependency.
- Enhances data privacy by keeping sensitive information on-device.
- Reduces latency for complex multi-step tasks by eliminating server round-trips.
Consensus
The model is a 24B parameter release specifically tuned for on-device agent performance.
Related articles (1)
Run fast, private agents on-device with LFM2.5-24B-A2BHugging Face Blog· 1 stories
More in Models
Google DeepMind Launches EmbeddingGemma 2 for Natively Multimodal Embeddings7 sources · Oct 13Meta Launches Muse Wonder and Wonder Pro Frontier Models2 sources · Oct 12Anthropic Releases Claude 5.5 Series Including Opus, Sonnet, and Haiku22 sources · Oct 7Mistral AI Launches 1-trillion Parameter Open-Weight Model Mistral Large 46 sources · Oct 7Strands Agents Releases Strands Decider 2b Decision Model1 source · Oct 7