---
format: "aidr-story-markdown/v1"
id: "0820332fc7e5d0ac603904d5c15cc728cfd4e6833c7a62fbeea7e05f7d86a1bf"
canonical_url: "https://aidr.today/0820332f?lang=en"
title: "Liquid AI Launches 24B Model as Largest Release to Date"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-10-13T17:57:59.000Z"
category: "Models"
topics: ["llm"]
source_urls: ["https://marketbrief.now/ai/liquid-ai-launches-24b-model-as-largest-release-to-date-60453b4d","https://x.com/liquidai/status/2110055807194931348","https://x.com/sgl_project/status/2110057968519856254","https://x.com/helloiamleonie/status/2110060663100117257","https://x.com/maximelabonne/status/2110058239312502889","https://huggingnews.com/ai/liquid-ai-launches-24b-model-as-largest-release-to-date-60453b4d"]
summary: "Liquid AI launched the LFM2.5-24B-A2B model, a 24B parameter reasoning tool using a Mixture-of-Experts architecture designed for fast, private agentic intelligence on laptops and PCs. The model features a 131K token context window, a 128K token vocabulary, and was trained on 33T tokens, with open base and post-trained weights available on Hugging Face. It reportedly outperforms Qwen3.6-35B-A3B on agentic tasks and reaches output speeds of 21K tokens per second on a single NVIDIA B300 and 184 tokens per second on an Apple M5 Max."
---

# Liquid AI Launches 24B Model as Largest Release to Date

> [Open the canonical story](<https://aidr.today/0820332f?lang=en>)

**Published:** 2026-10-13T17:57:59.000Z
**Category:** Models
**Topics:** llm

## Summary

Liquid AI launched the LFM2\.5\-24B\-A2B model, a 24B parameter reasoning tool using a Mixture\-of\-Experts architecture designed for fast, private agentic intelligence on laptops and PCs\. The model features a 131K token context window, a 128K token vocabulary, and was trained on 33T tokens, with open base and post\-trained weights available on Hugging Face\. It reportedly outperforms Qwen3\.6\-35B\-A3B on agentic tasks and reaches output speeds of 21K tokens per second on a single NVIDIA B300 and 184 tokens per second on an Apple M5 Max\.

## Sources

- [Story source](<https://marketbrief.now/ai/liquid-ai-launches-24b-model-as-largest-release-to-date-60453b4d>)
- [Story source](<https://x.com/liquidai/status/2110055807194931348>)
- [Supporting source](<https://x.com/sgl_project/status/2110057968519856254>)
- [Supporting source](<https://x.com/helloiamleonie/status/2110060663100117257>)
- [Supporting source](<https://x.com/maximelabonne/status/2110058239312502889>)
- [Story source](<https://huggingnews.com/ai/liquid-ai-launches-24b-model-as-largest-release-to-date-60453b4d>)