Liquid AI's LFM2.5-230M Beats Models Twice Its Size Running on a Raspberry Pi
Liquid AI's 230M-parameter LFM2.5 runs at 213 tok/s on a Galaxy S25 Ultra and beats models twice its size on tool use and data extraction
- New model: Liquid AI releases LFM2.5-230M, their smallest model yet, targeting edge and on-device agentic workloads.
- Speed: 213 tok/s on Galaxy S25 Ultra (CPU) and 42 tok/s on Raspberry Pi 5 — fastest in its class.
- Punches above weight: Beats models 2x+ its size on instruction following (IFEval: 71.71), tool use (BFCLv3: 43.26), and data extraction.
- Architecture: Hybrid of 8 gated short convolution blocks + 6 GQA attention blocks; pre-trained on 19T tokens with 32K context.
- Robot demo: Deployed on a Unitree G1 humanoid running on NVIDIA Jetson Orin, acting as a natural-language skill-selection layer.
- Availability: Free open-weight model on Hugging Face, with support for llama.cpp, MLX, vLLM, SGLang, and ONNX.
Liquid AI just pushed the floor on what a useful language model can look like. LFM2.5-230M is their smallest model yet: 230 million parameters that can run on a Raspberry Pi, a flagship Android phone, or a humanoid robot, all while outperforming models more than twice its size on the tasks that actually matter for agents and pipelines.
A different kind of small model
Most sub-500M models are curiosities. They're fine for demos but fall apart the moment you ask them to follow structured instructions, call tools reliably, or extract data from messy text. LFM2.5-230M competes with and often beats models more than twice as large, spanning instruction following, data extraction, and tool use. That's the headline claim, and the benchmarks back it up in the areas that count for agentic workloads.
The model scores 71.71 on IFEval (instruction following) and 43.26 on BFCLv3 (function calling), beating IBM's Granite 4.0-350M (53.48 and 39.58 respectively) and Gemma 3 1B (63.49 and 16.61) despite being a smaller model. On CaseReportBench, a data extraction benchmark, it scores 22.51 versus Gemma 3 1B's 2.28. These aren't marginal wins.
Where it does fall short is anywhere that requires deep reasoning. Given its compact size, Liquid does not recommend it for reasoning-heavy workloads such as advanced math, code generation, or creative writing. On MMLU-Pro, Qwen3.5-0.8B (a model more than 3x larger) scores 37.42 versus LFM2.5-230M's 20.25. The model knows what it is.
The architecture behind the speed
The speed story starts with the LFM2 architecture, which is neither a standard transformer nor a pure state-space model (SSM) like Mamba. The LFM2 architecture is a hybrid design that combines gated short convolutions for local sequence mixing and a minority of Grouped Query Attention (GQA) blocks for long-range token interaction. For the 230M variant specifically, the model has 14 layers: 8 double-gated LIV convolution blocks and 6 GQA blocks.