Liquid AI Releases LFM2.5-2.6B: On-Device Agentic Model for Edge Hardware

ArticleLiquid AIAnnouncement

Liquid AI has launched LFM2.5-2.6B, an open-weight 2.6-billion parameter model engineered for local agentic workflows on phones, CPUs, and enterprise GPUs.

Key points

  1. Features 2.6B parameters, pre-trained on roughly 34T tokens with a 128K context window and a doubled vocabulary for non-Latin scripts.
  2. Trained using a four-stage pipeline: two-stage SFT, specialized domain teachers via RLVR, Multi-Domain On-Policy Distillation (MOPD), and agentic RL using GRPO inside sandbox harnesses.
  3. Competes closely with or outperforms larger models such as Gemma-4 and Qwen3.5 across instruction following and tool-use evaluations, with larger models retaining an edge in code generation.
  4. Demonstrates high inference efficiency: 30 tokens/s on smartphones, 220 tokens/s on M5 Max, and 113 tokens/s on Ryzen AI Max+ 395 under 2.5 GB RAM, reaching nearly 15,000 tokens/s on an H100 GPU.
  5. Day-one support across inference runtimes including llama.cpp (GGUF), MLX, vLLM, SGLang, and ONNX, with compatibility for agent harnesses like Hermes Agent and OpenClaw.

Why it matters

Running agents on local hardware removes marginal per-token API costs, lowers latency, and ensures data privacy. This enables parallel background agentic operations and continuous multi-step task execution without relying on cloud infrastructure.

Analysis

LFM2.5-2.6B highlights the effectiveness of task-targeted distillation and multi-turn reinforcement learning inside actual runtime sandboxes. Liquid AI shows that small models can handle complex orchestration, tool calling, and long-context processing reliably. While cutting-edge coding still requires larger parameters, this release marks a significant step forward for privacy-first, edge-native agent deployment.

Mentioned links

Found in

Summary generated by AI (Gemini 3.8 Flash, 27 September 2026). It may contain errors – the original source is authoritative.

Related entries

Similar in content, found by the AI search.

  • Article:Chris Alexiuk and Chintan Patel

    NVIDIA Nemotron 3.5 Lightning: Compact MoE Model for Agentic Execution

    NVIDIA has announced Nemotron 3.5 Lightning, an open 30B mixture-of-experts (MoE) model featuring 3B active parameters. It is tailored for the high-volume execution layer of long-running AI agents, targeting routine tasks such as tool calling and result validation with low latency.

    AI & LLMs· Announcement

  • Article:Zhidx

    Ant Group open-sources compact MoE model Ling-3.0-tiny

    Ant Group's Bailian team has released Ling-3.0-tiny, an open-source lightweight mixture-of-experts model. It leverages DeepSeek's MLA and Kimi's KDA architectures with 7.9B total parameters, activating just 1.3B parameters per token for local edge deployment on hardware like Apple Silicon Macs and DGX systems.

    AI & LLMs· Announcement

  • Link:inclusionAI

    Ling-3.0-tiny: Lightweight Hybrid Reasoning MoE Model by inclusionAI

    inclusionAI has introduced Ling-3.0-tiny, a compact MoE model with 7.9B total parameters activating only 1.3B parameters per token. Utilizing a hybrid architecture of linear and latent attention mechanisms, it aims to deliver efficient reasoning and agentic workloads on local hardware.

    AI & LLMs· Tool

  • Repository:google-ai-edge/LiteRT-LM

    LiteRT-LM: Google's Open-Source Inference Framework for On-Device LLMs

    LiteRT-LM is an open-source inference orchestration layer from Google designed to run large language models directly on edge devices. It leverages GPU and NPU acceleration and powers on-device generative AI features in Chrome, Pixel Watch, and Chromebook Plus.

    6.4kstarsC++

    AI & LLMs· Tool

  • Link:Apodex AI

    Apodex-1.1-mini: Multimodal Agentic Model Released on Hugging Face

    Apodex AI has released Apodex-1.1-mini under the Apache-2.0 licence on Hugging Face. Based on the qwen3_5_moe architecture, the model is designed for long-horizon research and agentic workflows with native tool execution.

    AI & LLMs· Tool

Back to the knowledge base

Let's talk about your project

Get in touch

Locations

  • Mattersburg
    Johann Nepomuk Bergerstraße 7/2/14
    7210 Mattersburg, Austria
  • Vienna
    Ungargasse 64-66/3/404
    1030 Wien, Austria

Parts of this content were created with the assistance of AI.