AI · AI model
LFM2.5-1.2B-Thinking
Compact 1.2B reasoning model built for high-speed edge and browser inference.
LFM2.5-1.2B-Thinking is an edge-optimized hybrid reasoning model developed by Liquid AI that operates in under 1GB of memory. It delivers step-by-step reasoning for logic, math, and tool use with very fast local inference speeds on CPUs, mobile NPUs, and in browsers via WebGPU. It is designed specifically for local agentic workflows, data extraction, and offline client-side tasks.
What it's for
- Run client-side step-by-step reasoning directly in the browser via WebGPU without sending user data to servers
- Deploy agentic tool-use and structured data extraction pipelines on low-memory edge devices and laptops
- Power offline local chat assistants on mobile phones and embedded systems using llama.cpp or MLX
Features
- Sub-1GB memory footprint suitable for mobile NPUs, CPUs, and browser execution
- Hybrid architecture combining 10 double-gated convolution blocks with 6 grouped-query attention blocks
- 32,768 token context window with sustained decoding speeds across long contexts
- Native support for Pythonic and JSON tool calling in prompt templates
- Multilingual support covering 8 languages including English, Arabic, Chinese, French, German, Japanese, Korean, and Spanish
- Multi-format availability including Hugging Face native weights, GGUF, ONNX, and MLX
Pricing
Free model download under Liquid AI's LFM-1.0 custom license.
Details
- Type
- AI model
- Category
- AI
- Topic
- Small Language Models
- Open source
- No
- Platforms
- Web, Linux, macOS, Windows, Android, iOS
- For
- Machine learning engineers and web developers building private, offline, or resource-constrained AI applications.
- Alternative to
- Qwen3-1.7B, Llama 3.2 1B, Gemma 3 1B IT
- Tags
- edge ai, webgpu, reasoning, llm, slm, liquid ai
Why it was saved
- @HuggingModels: Private, offline-capable, no-install local inference at blazing speed; a reference point for running capable small reasoning models fully client-side.
Related
Open-weights AI models and chat interface for reasoning, coding, and research.
Conversational AI assistant with open-weights reasoning and general language models
Open-source reasoning model rivaling proprietary models in math, code, and logic
Conversational AI assistant for writing, research, reasoning, and coding
Browser that installs and runs open-source AI apps with one click
A comprehensive guide to prompt engineering techniques and LLM parameters by Google.
Updated