Workflows/Chat privately with LFM2.5-24B-A2B

Text / WORKFLOW

Chat privately with LFM2.5-24B-A2B

Run Liquid AI’s LFM2.5-24B-A2B with llama.cpp on Windows, Mac or Linux and chat in your browser, with every prompt staying on your computer.

Verified sourceSource checked 11/9/2026

LOCAL

Freshness 88 out of 100. Release age: 32 days old, Oct 13, 2026. Measures age, not quality.

Popularity 63 out of 100. 35,825 downloads in 30 days; 613 lifetime likes. Checked Nov 13, 2026. Measures adoption, not quality.

What do these scores mean?

Two ways to compare models. Neither score measures answer quality, benchmark performance, or whether a model fits your machine.

Freshness · release age

A new release starts at 100. The score halves every 180 days: 100 × 2^(−age in days / 180). When the release date is unavailable, a repository date is a fallback—not proof of when the model was released. Missing dates stay unscored.

Check the date source

Popularity · adoption

Downloads over the past 30 days carry 80% of the score; lifetime likes carry 20%. Both use a logarithmic scale, capped at 10 million downloads and 100,000 likes. If likes are unavailable, downloads carry all the weight. Missing download statistics stay unscored.

round(80 × min(1, log10(1 + downloads) / 7) + 20 × min(1, log10(1 + likes) / 5))

Check the statistics source

Scores as of Nov 14, 2026. Statistics checked Nov 13, 2026. Statistics are marked for refresh after 72 hours.Full scoring method

Before you begin

Hardware

Estimated from Liquid AI’s published file sizes plus runtime and context headroom. Only about 2.45B parameters are active per token, so partial offload to system RAM stays usable but slower. Not measured by YouRunAI.

Machine	LFM2.5-24B-A2B · Q4_K_M GGUF	LFM2.5-24B-A2B · Q8_0 GGUF
NVIDIA · 16 GB VRAM + 32 GB RAM	Runs with some layers offloaded to RAM; slower but usable	Heavy offload; choose Q4_K_M instead
NVIDIA RTX 4090 · 24 GB VRAM	Recommended start; fits on the GPU with the starting context	Does not fit fully; offload to RAM or use Q4_K_M
NVIDIA RTX 5090 · 32 GB VRAM	Comfortable, with room for longer context	Fits at the starting context; check free VRAM
AMD Radeon · 24 GB VRAM	Fits; requires the HIP or Vulkan llama.cpp build on Windows (Linux tab here is NVIDIA only)	Partial offload; requires the HIP or Vulkan build
Intel Arc · 16 GB VRAM + 32 GB RAM	Partial offload through the Vulkan llama.cpp build; slower	Heavy offload; choose Q4_K_M instead
Apple Silicon · 24 GB unified memory	Tight: close other apps and keep the starting context	Cannot fit; use Q4_K_M or a larger Mac
Apple Silicon · 32 GB unified memory	Recommended Mac start	Cannot fit in the default GPU memory budget; use Q4_K_M
Apple Silicon · 64 GB unified memory	Comfortable	Comfortable at the starting context
CPU only · 32 GB RAM	Runs on CPU; slow but usable for short tasks	Needs 48 GB RAM or more; use Q4_K_M

Both packages are Liquid AI’s own GGUF files. Liquid AI also publishes a 13.6 GB MLX 4-bit build for Apple Silicon; this guide uses llama.cpp on every platform so the same server works for chat, Cline and Hermes.

PRIVATE CHAT · WINDOWS · MAC · LINUXSOURCE-LINKED SETUP

Chat privately with LFM2.5-24B-A2B

Download Liquid AI’s GGUF once, start llama.cpp’s local server, and chat in your browser. The model reasons before every answer and keeps context across follow-ups.

Choose a package Run its commands Check the result

BEFORE YOU BEGIN

Choose your package

Select the weights you want. The instructions, downloads and settings change with your choice.

The publisher has not provided a complete memory target for this task. Check its hardware guidance before starting.

CURRENT SETUP

LFM2.5-24B-A2B · Q4_K_M GGUF · official

Start at 16,384 tokens for this chat setup. The 14.5 GB file is only the weights; the runtime and context add more memory. Longer context needs more memory, so reduce it first if the server runs out. Estimate, not a YouRunAI measurement.

Freshness 88 out of 100. Release age: 32 days old, Oct 13, 2026. Measures age, not quality.

Popularity 63 out of 100. 35,825 downloads in 30 days; 613 lifetime likes. Checked Nov 13, 2026. Measures adoption, not quality.

What do these scores mean?

Two ways to compare models. Neither score measures answer quality, benchmark performance, or whether a model fits your machine.

Freshness · release age

Check the date source

Popularity · adoption

round(80 × min(1, log10(1 + downloads) / 7) + 20 × min(1, log10(1 + likes) / 5))

Check the statistics source

Scores as of Nov 14, 2026. Statistics checked Nov 13, 2026. Statistics are marked for refresh after 72 hours.Full scoring method

View weight source

THE WORKFLOW

Set up LFM2.5-24B-A2B · Q4_K_M GGUF on your machine

Choose your computer first, then its GPU path. The instructions below update to match your selection.

CHOOSE YOUR SETUPWindows

Install llama.cpp with winget

Open PowerShell and install the official llama.cpp release with winget, then close and reopen PowerShell so llama-server is on your PATH. The second command lists the compute devices llama.cpp can use. If only the CPU appears on an NVIDIA or AMD PC, download the CUDA or HIP Windows build from the official releases page instead, extract it, and run the same commands from that folder.

Official llama.cpp releases (Windows builds) Liquid AI GGUF package and run snippets

PowerShell

winget install llama.cpp
llama-server --list-devices

Download Q4_K_M and start the local server

In PowerShell, run this command and leave the window open. The first run downloads the 14.5 GB Q4_K_M file from Liquid AI's GGUF repository; later runs reuse it. The server listens only on this computer at http://127.0.0.1:8080, names the model lfm2.5-24b, and starts with a 16,384-token context and Liquid AI's recommended sampling (temperature 0.1, top_k 50, repetition penalty 1.05). Wait until the log says the server is listening.

Exact Q4_K_M file in the publisher repository llama.cpp server options

PowerShell

llama-server -hf LiquidAI/LFM2.5-24B-A2B-GGUF:Q4_K_M --alias lfm2.5-24b --jinja -c 16384 --temp 0.1 --top-k 50 --repeat-penalty 1.05 --host 127.0.0.1 --port 8080

Open the chat page and send a first message

Open http://127.0.0.1:8080 in your browser on the same computer. llama.cpp serves its own chat interface there. Start a new conversation and send the prompt below. LFM2.5-24B-A2B always reasons before it answers, so you will first see a collapsible thinking section and then the reply.

Plain text

Plan a three-day study schedule for learning basic Spanish. Use one short bullet per day.

Ask a follow-up that depends on the first answer

In the same conversation, send: “Make day two lighter and move the vocabulary drill to day three.” A correct reply edits the plan it already gave you rather than starting over. Liquid AI notes that the model is not meant for knowledge-heavy questions on its own, so check facts and use it for planning, rewriting and reasoning tasks.

Use it privately, stop and restart

Once the file has downloaded, the conversation runs entirely on this computer; you can disconnect from the internet and send a new prompt to confirm. Chats are kept in this browser's storage, so use the page's delete controls on a shared computer. Press Ctrl+C in PowerShell to stop the server. To chat again, run the same command and reopen http://127.0.0.1:8080; no second download is needed.

SUCCESS CHECK · LFM2.5-24B-A2B · Q4_K_M GGUF

A private conversation that remembers context

The chat page at http://127.0.0.1:8080 answers your first prompt, then revises its own plan when you ask the follow-up, with the internet disconnected.

http://127.0.0.1:8080Opens on your machine

Sources checked November 9, 2026. Commands follow Liquid AI’s published GGUF and settings; YouRunAI has not run this exact setup. Hardware ratings are estimates.

Back to setup steps

When it doesn’t go to plan

The server stops with an out-of-memory error.

Lower -c to 8192, close other GPU apps, or switch to Q4_K_M. On a 16 GB GPU add --n-gpu-layers 30 to keep part of the model in system RAM.

llama-server reports an unknown architecture.

Your llama.cpp is too old for lfm2moe models. Update it (winget upgrade llama.cpp, brew upgrade llama.cpp, or docker pull the server-cuda image) and try again.

Answers start with a long thinking section.

That is expected: LFM2.5-24B-A2B has no non-thinking mode. Collapse the thinking section in the chat page and read the final answer.

REFERENCE LIBRARY

Sources and files

Original instructions, model files, and compatibility notes behind this setup.

4 SOURCES

01Liquid AI model card, license and recommended settingshuggingface.co 02Official LFM2.5-24B-A2B GGUF fileshuggingface.co 03llama.cpp releases and servergithub.com 04LFM Open License v1.0huggingface.co

The model behind this workflow

LFM2.5-24B-A2B

Keep exploring

Text

SETUP GUIDE

Qwen3.5 · 9B private chat setup

Source-backed package and platform-route plan for private local chat with Qwen3.5 · 9B. Preserves the official Ollama 9B package, a smaller 4B family fallback, and a clearly labeled community uncensored GGUF option.

Freshness 37 out of 100. Release age: 260 days old, Feb 27, 2026. Measures age, not quality.

Build a private coding assistant with Qwen3.8 27B

Run Qwen3.8 27B with Ollama, connect it to Cline in VS Code, and make one reviewable code change in a tiny local project.

Freshness 68 out of 100. Repository age: 101 days old, Aug 5, 2026. Measures age, not quality.

Arena Open Source Web coding: rank 13 of 63. Ranked model: Qwen3.8-27B. Original model ranking, not a benchmark of this package. Export dated Nov 11, 2026. Source checked Nov 13, 2026. Human preference ranking, not hardware suitability.+3

SOFTWAREOllama + Visual Studio Code + Cline + Python

Create images locally with FLUX.2 Klein

Generate images from text prompts in ComfyUI with the fast distilled or more adjustable base 4B model. Save your results and reuse the visual workflow.

Freshness 31 out of 100. Release age: 303 days old, Jan 15, 2026. Measures age, not quality.

SOFTWAREComfyUI

LOCALRENTED GPU

Verified source