Liquid AI Releases LFM2.5-2.6B: An On-Device Agentic Model With 128K Context, Tool Calling, And Open Weights
Liquid AI launched LFM2.5-2.6B, a 2.69B parameter on-device agentic model supporting 128K context and tool calling with open weights.
- LFM2.5-2.6B is a 2.69B parameter on-device agentic model with a 128K context window.
- The model supports tool calling for multi-step tasks and runs efficiently on consumer hardware.
- Open weights are available in GGUF, MLX, and ONNX formats for broad accessibility.
- Achieves 220 tokens/sec on an M5 Max with under 2.5 GB memory usage.
Liquid AI has introduced LFM2.5-2.6B, a compact agentic model designed to run entirely on-device. The model features 2.69 billion parameters and supports a context window of 131,072 tokens, allowing it to process long documents or conversations without external dependencies. It includes built-in tool-calling capabilities, enabling multi-step task execution such as web searches or API interactions directly on the device.
The architecture combines 22 double-gated short convolution blocks with 8 grouped-query attention (GQA) blocks across 30 layers, optimizing for both performance and efficiency. On an M5 Max, the model achieves 220 tokens per second with under 2.5 GB of memory usage, making it feasible for deployment on consumer hardware. Liquid AI has released the model under open weights in GGUF, MLX, and ONNX formats, facilitating broad accessibility for developers and researchers.
This release underscores Liquid AI’s focus on enabling private, low-latency AI applications that do not require cloud connectivity. The open-weight approach aligns with growing industry demand for transparent and customizable AI solutions.
Provides a lightweight, open-weight model for on-device agentic applications with tool-calling support.
Enables private, low-latency AI solutions without cloud dependency, reducing operational costs.
Highlights a niche in efficient, on-device AI models with growing market potential.
Demonstrates practical applications of agentic models and efficient model architectures.
- agentic model
- An AI model capable of planning, executing multi-step tasks, and interacting with tools or APIs autonomously.
- GGUF
- A file format for quantized large language models, optimized for efficient inference.
- MLX
- Apple's machine learning framework for efficient model deployment on Apple silicon.
AI ToolsAIOps Agents for Kubernetes Human-in-the-Loop Remediation on GCP
AI ToolsBackflip AI turns 3D scans into editable CAD models in minutes instead of hours
AI ToolsAngular WebMCP — Your App is Now an AI Tool 🔥🚀
AI ToolsxAI's Imagine Image 2.0 lands just behind OpenAI's GPT-Image-2 in Arena benchmarks
AI ToolsMistral AI Releases Shieldstral 1.0 3B: An Open-Weights Policy-Adaptive Multimodal Safety Classifier Matching Models 7× Its Size
Military use, cyberattacks and political messaging: How governments around the world are approaching AI - i24NEWS
Governments worldwide are exploring AI for military use, cyberattacks, and political messaging. This trend is driven by the need to stay ahead in terms of technology and security.
A law promoted by Gavin Newsom is now in effect in California, changing how content created with artificial intelligence is identified - Clarin.com
A new law in California, promoted by Governor Gavin Newsom, requires content creators to identify AI-generated content. This change aims to increase transparency and accountability.
Iran wins medals at International Olympiad in Artificial Intelligence - Tehran Times
Iranian students won medals at the International Olympiad in Artificial Intelligence, highlighting the country's growing AI education and research capabilities.
RoboticsThe first self-driving vehicle on Mars has proven to be a smashing success
NASA's Perseverance rover has autonomously driven nearly 90% of its total distance on Mars, demonstrating a major leap in planetary robotic autonomy.
AI ResearchFields Medalist who published a paper on AI-driven human extinction now works for OpenAI
Fields Medalist Jacob Tsimerman is leaving academia to join OpenAI, where he will work on AI safety after publishing a paper on AI-driven human extinction risks.
AI ResearchDeepMind’s hurricane breakthrough has surprised weather scientists
DeepMind’s new open-source WeatherNext model delivers accurate hurricane predictions using lower-resolution weather data, giving forecasters an extra day of lead time.