---
source: 'https://howaiworks.ai/blog/liquid-ai-lfm2-5-thinking-on-device'
section: blog
title: 'Liquid AI LFM2.5-1.2B-Thinking: On-Device Reasoning Under 1GB'
description: >-
  Liquid AI LFM2.5-1.2B-Thinking is a breakthrough 1.2B reasoning model that
  fits in 900MB, delivering high-performance logic on phones and laptops.
date: '2026-03-03'
lastUpdated: '2026-07-21'
author: HowAIWorks Team
tags:
  - Liquid AI
  - LFM 2.5
  - On-Device AI
  - Edge AI
  - Reasoning Models
  - Machine Learning
  - NPU
  - AMD Ryzen
  - Qualcomm
  - Thinking Models
readingTime: 6 minutes
isReviewed: true
---

# Liquid AI LFM2.5-1.2B-Thinking: On-Device Reasoning Under 1GB

> Liquid AI LFM2.5-1.2B-Thinking is a breakthrough 1.2B reasoning model that fits in 900MB, delivering high-performance logic on phones and laptops.

## Introduction

The promise of true "thinking" AI—models that don't just predict the next word but internalize logical steps before speaking—has traditionally been the domain of massive data centers. Running these reasoning-heavy workloads usually requires high-end GPUs and significant power consumption. However, [Liquid AI](https://www.liquid.ai/) has just shattered this barrier with the release of **LFM2.5-1.2B-Thinking**.

This new addition to the [LFM 2.5 family](https://howaiworks.ai/blog/liquidai-lfm2-5-1-2b-thinking-release) is a 1.2 billion parameter model specifically engineered for sophisticated reasoning tasks on [edge devices](https://howaiworks.ai/glossary/edge-ai). What required a cluster of servers two years ago now runs offline, entirely on-device, fitting within a mere 900 MB of memory. This release signals a critical shift toward "Reasoning-as-a-Service" on hardware as small as a smartphone or a laptop.

## The Power of Internal Reasoning Traces

Unlike standard instruction-following models that provide immediate answers, LFM2.5-1.2B-Thinking generates **thinking traces**. This process allows the model to work through problems systematically, verifying intermediate results and adjusting its approach before producing the final output.

The impact of this "reason-first" architecture is most visible in complex, objective benchmarks:

- **Mathematical Reasoning**: On the MATH-500 benchmark, the model's score skyrocketed from **63 to 88** compared to the LFM2.5-1.2B-Instruct version.
- **Instruction Following**: Improved from **61 to 69** on the Multi-IF benchmark.
- **Agentic Tool Use**: Jumped from **49 to 57** on BFCLv3 (Berkeley Function Calling Leaderboard).

These metrics demonstrate that even at a compact 1.2B scale, a model can exhibit high-level logic if trained with the right "thinking" recipe.

## Technical Breakthroughs: The Training Recipe

Building a capable reasoning model at such a small scale presented unique challenges, particularly the "knowledge capacity" limit. Liquid AI addressed this through a sophisticated training and alignment pipeline:

### Eliminating the "Doom Loop"

A common failure mode for small reasoning models is the **"doom loop"**—getting stuck in repetitive text patterns instead of reaching a conclusion. Liquid AI mitigated this during preference alignment by sampling multiple candidates (5 temperature-sampled and 1 greedy) and using an LLM judge to favor non-looping, high-quality responses.

### RLVR and GRPO-style Optimization

The model's reinforcement learning (RL) pipeline is built on a fork of *verl*, focusing on **critic-free, group-relative policy-gradient optimization (GRPO-style)**. By applying techniques like asymmetric ratio clipping and dynamic filtering, the team reduced the occurrence of doom loops from **15.74% in mid-training to just 0.36%** in the final RLVR (Reinforcement Learning from Verifiable Rewards) stage.

### Curriculum RL and Model Merging

Rather than training a single monolithic model for all tasks, Liquid AI used a **Curriculum RL** approach. They created domain-specific checkpoints for instruction following, math, and tool use, and then used **iterative model merging** to combine these specialized capabilities. This flexibility allowed them to navigate trade-offs more effectively than sequential training.

## Performance and Hardware Ecosystem

The LFM2.5-1.2B-Thinking model isn't just a research experiment; it's ready for deployment. Liquid AI has expanded its partner ecosystem to ensure high-performance execution across diverse hardware, and in particular across the [NPUs](https://howaiworks.ai/glossary/neural-processing-unit) now shipping in laptops and phones:

- **AMD Ryzen NPUs**: Using the specialized *FastFlowLM* runtime, the model sustains **52 tokens per second** at 16K context and **46 tokens per second** even at its full 32K context.
- **Qualcomm NPUs**: Optimized through a partnership with *Nexa AI*, bringing fast inference to mobile and PC devices.
- **Cross-Platform Support**: "Day-zero" support is available for *llama.cpp*, *MLX*, *vLLM*, and *ONNX Runtime*, covering Apple, AMD, Qualcomm, and NVIDIA hardware.

> [!TIP]
> While the Thinking model excels at logic and agentic tasks, Liquid AI recommends using the [LFM2.5-1.2B-Instruct](https://howaiworks.ai/blog/liquidai-lfm2-5-1-2b-thinking-release) for standard chat and creative writing, where reasoning traces might add unnecessary latency.

## Conclusion

The release of LFM2.5-1.2B-Thinking proves that **size is not the only factor in intelligence**. By optimizing architecture and training recipes for reasoning rather than just raw scale, Liquid AI is enabling a new generation of [AI agents](https://howaiworks.ai/glossary/ai-agent) that can think, plan, and use tools entirely in your pocket. As on-device AI matures, these compact "thinking" models will become the backbone of privacy-first, ultra-fast digital assistants.

_To stay updated on the latest AI developments, explore our [AI models catalog](https://howaiworks.ai/models), check out our [glossary](https://howaiworks.ai/glossary) for key terms, or browse our [blog](https://howaiworks.ai/blog) for more insights._

## Sources

- [Liquid AI Official Blog Announcement](https://www.liquid.ai/blog/lfm2-5-1-2b-thinking-on-device-reasoning-under-1gb)
- [LiquidAI/LFM2.5-1.2B-Thinking on Hugging Face](https://huggingface.co/LiquidAI/LFM2.5-1.2B-Thinking)
- [FastFlowLM AMD Ryzen Benchmarks](https://fastflowlm.com/docs/benchmarks/lfm2_results/)
- [LFM2.5 Family Models and Playgrounds](https://playground.liquid.ai/)

## Frequently Asked Questions

### How does LFM2.5-1.2B-Thinking differ from the Instruct version?

The Thinking version generates internal reasoning traces before answering, significantly improving performance on math, logic, and tool-use tasks (e.g., MATH-500 score jumped from 63 to 88).

### What is the memory footprint of this model?

The model is designed to run entirely on-device, fitting within approximately 900MB of memory, making it ideal for mobile phones and laptops.

### Which hardware platforms are supported at launch?

It supports Apple, AMD, Qualcomm, and NVIDIA hardware through frameworks like llama.cpp, MLX, vLLM, and ONNX Runtime, with specific optimizations for AMD Ryzen and Qualcomm NPUs.

### Does the model suffer from 'doom looping'?

Liquid AI used a specialized RLVR (Reinforcement Learning from Verifiable Rewards) pipeline to reduce 'doom loops' (repetitive text patterns) from 15.74% down to 0.36%.

### Is this model better than Qwen3-1.7B?

Benchmarks show it matches or exceeds Qwen3-1.7B on most reasoning tasks despite having 40% fewer parameters, offering better efficiency in terms of output tokens and speed.

---

Source: https://howaiworks.ai/blog/liquid-ai-lfm2-5-thinking-on-device — HowAIWorks.ai