---
title: LFM2-24B-A2B — Efficient MoE LLM | ModelsLab
description: Run LiquidAI LFM2-24B-A2B hybrid MoE model with 2.3B active params for fast agentic inference. Try high-throughput RAG and multi-turn tasks now.
url: https://modelslab-frontend-v2-927501783998.us-east4.run.app/liquidai-lfm2-24b-a2b
canonical: https://modelslab-frontend-v2-927501783998.us-east4.run.app/liquidai-lfm2-24b-a2b
type: website
component: Seo/ModelPage
generated_at: 2026-05-13T09:42:51.507971Z
---

Available now on ModelsLab · Language Model

LiquidAI: LFM2-24B-A2B
Fast MoE Inference Engine
---

[Try LiquidAI: LFM2-24B-A2B](/models/open_router/liquid-lfm-2-24b-a2b) [API Documentation](https://docs.modelslab.com)

Scale Agents Efficiently
---

Hybrid MoE

### 24B Params 2.3B Active

Activates 2.3B params per token in 40-layer A2B architecture with 30 conv blocks.

Low Memory

### Fits 32GB RAM

Deploys on laptops, edge devices, and H100s for LiquidAI: LFM2-24B-A2B API workflows.

High Throughput

### 26K Tokens Second

Handles 1024 concurrent requests at 32K context for liquidai lfm2 24b a2b pipelines.

Examples

See what LiquidAI: LFM2-24B-A2B can create
---

Copy any prompt below and try it yourself in the [playground](/models/open_router/liquid-lfm-2-24b-a2b).

Math Proof

“Prove the Pythagorean theorem step-by-step using geometric arguments and formal logic. Include diagrams in ASCII art and verify with coordinates.”

Code Debugger

“Analyze this Python function for bugs: def factorial(n): if n == 0: return 1 else: return n \* factorial(n-1). Fix recursion depth issues and optimize for large n.”

Agent Workflow

“Plan a multi-step research task: query database for sales data, analyze trends with stats, generate report in JSON, and suggest actions.”

RAG Summary

“Summarize key insights from these documents on climate models, extract trends, and output structured JSON with citations for 32K context.”

For Developers

A few lines of code.
Agents. Two Lines.
---

ModelsLab handles the infrastructure: fast inference, auto-scaling, and a developer-friendly API. No GPU management needed.

- **Serverless:** scales to zero, scales to millions
- **Pay per token,** no minimums
- **Python and JavaScript SDKs,** plus REST API

[API Documentation ](https://docs.modelslab.com)

PythonJavaScriptcURL

Copy

```
<code>import requests

response = requests.post(
    "https://modelslab.com/api/v7/llm/chat/completions",
    json={
  "key": "YOUR_API_KEY",
  "prompt": "",
  "model_id": ""
}
)
print(response.json())</code>
```

FAQ

Common questions about LiquidAI: LFM2-24B-A2B
---

[Read the docs ](https://docs.modelslab.com)

### What is LiquidAI: LFM2-24B-A2B?

Hybrid MoE LLM with 24B total params, 2.3B active per token. Uses 30 gated conv + 10 GQA blocks in A2B ratio. Fits 32GB RAM for edge deployment.

### How fast is liquidai lfm2 24b a2b?

Reaches 26.8K tokens/sec on H100 at 1024 requests. Outperforms Qwen3-30B-A3B in throughput. Supports 32K context for agents.

### What is LiquidAI: LFM2-24B-A2B API used for?

Powers high-volume multi-agent pipelines, RAG, function calling. Excels in instruction following, math reasoning, tool use. Native web search support.

### Is LiquidAI: LFM2-24B-A2B model efficient?

Activates sparse 2.3B params for low memory cost. Runs on consumer laptops with iGPU/NPU. 112 tok/s on AMD CPU.

### LiquidAI: LFM2-24B-A2B alternative to what?

Beats gpt-oss-20b, Qwen3-30B-A3B in benchmarks. Ideal dense model replacement for agentic tasks. Scales from 350M to 24B.

### Where to access liquidai lfm2 24b a2b api?

Available via Together AI, LM Studio, OpenVINO. Supports vLLM, llama.cpp. Deploy locally or cloud for privacy.

Ready to create?
---

Start generating with LiquidAI: LFM2-24B-A2B on ModelsLab.

[Try LiquidAI: LFM2-24B-A2B](/models/open_router/liquid-lfm-2-24b-a2b) [API Documentation](https://docs.modelslab.com)

---

*This markdown version is optimized for AI agents and LLMs.*

**Links:**
- [Website](https://modelslab.com)
- [API Documentation](https://docs.modelslab.com)
- [Blog](https://modelslab.com/blog)

---
*Generated by ModelsLab - 2026-05-13*