Mu Description
On June 23, 2025, Microsoft unveiled Mu, an innovative 330-million-parameter encoder–decoder language model specifically crafted to enhance the agent experience within Windows environments by effectively translating natural language inquiries into function calls for Settings, all processed on-device via NPUs at a remarkable speed of over 100 tokens per second while ensuring impressive accuracy. By leveraging Phi Silica optimizations, Mu’s encoder–decoder design employs a fixed-length latent representation that significantly reduces both computational demands and memory usage, achieving a 47 percent reduction in first-token latency and a decoding speed that is 4.7 times greater on Qualcomm Hexagon NPUs when compared to other decoder-only models. Additionally, the model benefits from hardware-aware tuning techniques, which include a thoughtful 2/3–1/3 split of encoder and decoder parameters, shared weights for input and output embeddings, Dual LayerNorm, rotary positional embeddings, and grouped-query attention, allowing for swift inference rates exceeding 200 tokens per second on devices such as the Surface Laptop 7, along with sub-500 ms response times for settings-related queries. This combination of features positions Mu as a groundbreaking advancement in on-device language processing capabilities.
Mu Alternatives
Qloo
Qloo, the "Cultural AI", is capable of decoding and forecasting consumer tastes around the world. Privacy-first API that predicts global consumer preferences, catalogs hundreds of million of cultural entities, and is privacy-first. Our API provides contextualized personalization and insight based on deep understanding of consumer behavior. We have access to more than 575,000,000 people, places, and things. Our technology allows you to see beyond trends and discover the connections that underlie people's tastes in their world. Our vast library includes entities such as brands, music, film and fashion. We also have information about notable people. Results are delivered in milliseconds. They can be weighted with factors like regionalization and real time popularity. Companies who want to use best-in-class data to enhance their customer experiences. Our flagship recommendation API provides results based on demographics and preferences, cultural entities, metadata, geolocational factors, and metadata.
Learn more
Vehicle Acquisition Network (VAN)
Vehicle Acquisition Network (VAN) is an automotive software platform built to help car dealerships acquire high-quality used vehicles directly from private sellers—without relying on auctions. As wholesale prices rise and vehicle availability tightens, VAN empowers dealers with tools to source inventory faster, more profitably, and with greater control.
VAN aggregates local FSBO (for-sale-by-owner) listings, applies real-time market data to assess profitability, and automates communication with sellers at scale. Buyers can manage leads, track seller conversations, and streamline acquisition workflows through an intuitive CRM-style dashboard designed specifically for dealership teams.
For dealers who don’t have dedicated acquisition staff, VAN offers a Managed Buyer program, pairing stores with expert buyers who actively source, engage, and negotiate with private sellers on their behalf—saving time and boosting acquisition volume without internal hiring.
VAN is trusted by hundreds of dealerships across North America—from independent rooftops to franchise groups—looking to beat Carvana and CarMax at their own game. It's the smarter way to buy cars.
Learn more
Whisper
We have developed and are releasing an open-source neural network named Whisper, which achieves levels of accuracy and resilience in English speech recognition that are comparable to human performance. This automatic speech recognition (ASR) system is trained on an extensive dataset comprising 680,000 hours of multilingual and multitask supervised information gathered from online sources. Our research demonstrates that leveraging such a comprehensive and varied dataset significantly enhances the system's capability to handle different accents, ambient noise, and specialized terminology. Additionally, Whisper facilitates transcription across various languages and provides translation into English from those languages. We are making available both the models and the inference code to support the development of practical applications and to encourage further exploration in the field of robust speech processing. The architecture of Whisper follows a straightforward end-to-end design, utilizing an encoder-decoder Transformer framework. The process begins with dividing the input audio into 30-second segments, which are then transformed into log-Mel spectrograms before being input into the encoder. By making this technology accessible, we aim to foster innovation in speech recognition technologies.
Learn more
CodeT5
CodeT5 is an innovative pre-trained encoder-decoder model specifically designed for understanding and generating code. This model is identifier-aware and serves as a unified framework for various coding tasks. The official PyTorch implementation originates from a research paper presented at EMNLP 2021 by Salesforce Research. A notable variant, CodeT5-large-ntp-py, has been fine-tuned to excel in Python code generation, forming the core of our CodeRL approach and achieving groundbreaking results in the APPS Python competition-level program synthesis benchmark. This repository includes the necessary code for replicating the experiments conducted with CodeT5. Pre-trained on an extensive dataset of 8.35 million functions across eight programming languages—namely Python, Java, JavaScript, PHP, Ruby, Go, C, and C#—CodeT5 has demonstrated exceptional performance, attaining state-of-the-art results across 14 different sub-tasks in the code intelligence benchmark known as CodeXGLUE. Furthermore, it is capable of generating code directly from natural language descriptions, showcasing its versatility and effectiveness in coding applications.
Learn more
Integrations
No Integrations at this time
Company Details
Company:
Microsoft
Year Founded:
1975
Headquarters:
United States
Website:
blogs.windows.com/windowsexperience/2025/06/23/introducing-mu-language-model-and-how-it-enabled-the-agent-in-windows-settings/
Media
Recommended Products
Auth0 B2B Essentials: SSO, MFA, and RBAC Built In
Auth0's B2B Essentials plan gives you everything you need to ship secure multi-tenant apps. Unlimited orgs, enterprise SSO, RBAC, audit log streaming, and higher auth and API limits included. Add on M2M tokens, enterprise MFA, or additional SSO connections as you scale.
Product Details
Platforms
Windows
Types of Training
Training Docs
Live Training (Online)
Webinars
In Person
Training Videos
Customer Support
Business Hours
Online Support
Mu Features and Options
Mu Lists
Mu User Reviews
Write a Review- Previous
- Next
