AI model · by Liquid AI
LFM2.5-24B-A2B
LFM2.5-24B-A2B is Liquid AI's 24B-parameter reasoning model, with just 2B active, made for speedy, private agents on local devices.
- 128K tokens context
- Open weights · Lfm1.0 licence
- Checked against Liquid AI's page 19 Oct 2026
What LFM2.5-24B-A2B is for
Liquid AI says this reasoning model has 24B parameters but activates just 2B of them, aiming it at quick agents that keep data on the device. It supports tool calling for agentic workflows and handles a 131,072-token context length. Liquid AI also advises against using it alone for knowledge-intensive work and recommends pairing it with retrieval or web search.
What Liquid AI says it does well
- Competitive with Qwen3.6-35B-A3B on agentic and tool-use tasks.
- Always thinks before it answers, with cheap reasoning tokens due to 2B active parameters per token.
- Runs up to 1.7× faster than comparable MoEs on-device.
- Supports tool calling for agentic workflows.
- Context length of 131,072 tokens.
Limits Liquid AI mentions
- Not recommended for knowledge-heavy tasks on its own; should be paired with retrieval or web search.
- Text-only model.
- Unlike LFM2-24B-A2B, it is a pure reasoning model that always thinks before it answers.