Liquid AI ships LFM2.5-24B-A2B open weights for local agents

A 2B-active reasoning model with open base and post-trained weights targets multi-step agent work on laptops and PCs, expanding what can run locally in agent harnesses.

• Liquid AI's LFM2.5-24B-A2B is a reasoning model with 24B total parameters but only 2B active per token, aimed at running multi-step agent tasks on laptops and PCs.
• Liquid lists a 131K-token context window, a 128K-token vocabulary and 33T pretraining tokens, and releases both base and post-trained weights on Hugging Face.
• Liquid claims it performs competitively on agentic and tool-use tasks with Qwen3.6-35B-A3B, which has roughly 50% more total and active parameters; the post cites no specific scores.

• context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
• tool use — A model's ability to call external functions — run code, search the web, edit files — instead of only generating text.
• reasoning model — A model trained to think — generating extended internal reasoning before answering — trading time and tokens for accuracy on hard problems.
• pretraining — The first, biggest phase of building a model: training it on enormous amounts of text so it learns language, facts, and reasoning in general.

• postToday we release Open d1: two open-weight multimodal models in our d1 decision…
• postLiquid AI's d1 decision model adds vision at a fraction of frontier cost
• articled1 is now available on @vercel AI gateway!We worked with the amazing vercel…
• postLiquid AI ships d1, a decision model for structured software tasks

• postLFM2.5-2.6B: an open-weight agentic model that plans and calls tools on-deviceLiquid AI
• articleRun fast, private agents on-device with LFM2.5-24B-A2Bhuggingface.co
• postWe're releasing HY-Embodied-0.5, a family of foundation models for real-world…Tencent Hy

Checking sign-in…

Loading comments…