● Covered by 9 sources · 15 reportsMedium impact15 neutral
Meta Releases Muse Code AI Coding Agent and Open-Source Muse Glimmer Model
Key points
- Muse Code is a terminal AI coding agent for macOS/Linux.
- It is powered by the new Muse Spark 1.2 model.
- Muse Glimmer is a 30B parameter open-weight model for local AI agents.
- Muse Glimmer is licensed under Apache 2.0 and runs on consumer GPUs.
- Meta positions Muse Code as a more affordable alternative to competitors.
Meta Enters AI Coding Agent Market with Muse Code
Meta has released Muse Code, a new terminal-based AI coding agent available in beta for macOS and Linux. This tool is designed to handle complex software engineering tasks across large code repositories, including planning changes, writing code, and validating results. It operates with a simple agent loop and specialized asynchronous background agents that remain active throughout a session, reducing latency and the need for user intervention.
Muse Code is powered by Meta's new Muse Spark 1.2 model, which offers improvements in code generation, debugging, codebase understanding, and long-running developer workflows. The model is accessible through Muse Code and the Meta Model API.
Competitive Positioning and Pricing Strategy
The launch of Muse Code positions Meta in direct competition with existing AI coding tools from companies like Anthropic (Claude Code) and OpenAI (Codex). Meta CEO Mark Zuckerberg stated that Muse Code can accomplish "complete software engineering tasks across large repos." Meta's AI chief Alexandr Wang indicated that Muse Code aims to be more affordable than its competitors.
The standard Muse Spark 1.2 model is priced at $1.25 per million input tokens and $4.25 per million output tokens. A "contributor" tier is available at $0.10 and $0.20 per million tokens, which is significantly cheaper. However, this lower price requires users to opt-in to share their code, prompts, and sessions for Meta's product improvement, raising data privacy concerns for some engineering leaders.
Introduction of Muse Glimmer for Local AI
Alongside Muse Code, Meta also released Muse Glimmer, a 30-billion-parameter open-weight AI model. Muse Glimmer is designed to run autonomous AI agents directly on consumer hardware, such as Macs and PCs with a single GPU, without relying on cloud infrastructure. This model is based on Meta's Muse Spark 1.2 model and is released under the permissive Apache 2.0 license, allowing for unrestricted commercial use, modification, and redistribution.
Muse Glimmer supports text and images and was trained across over 100 languages. It is intended for agent-oriented tasks like scheduling, file management, and multi-step workflows, enabling privacy-sensitive personal AI agents by processing data locally.
Distillation and Open-Weight Strategy
Muse Glimmer was developed using a multi-stage training strategy, including logit distillation from Muse Spark, mid-training on long-context sequences, and post-training alignment through supervised fine-tuning and reinforcement learning. This approach demonstrates how larger cloud-based models can be distilled into smaller, locally deployable agents.
Meta's release of Muse Glimmer and its commitment to opening the weights for Muse Spark 1.2 signal a renewed focus on open-weight models, differentiating its AI strategy from companies that develop proprietary models.
✨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors — check the original sources. How BrevFeed works →
The daily brief
One email each morning: the day's tech stories, clustered across outlets and summarized. No account needed.
One email a day. Unsubscribe in one click, any time.
Today's brief
Spend a few minutes, get the whole day. Every topic's top stories in one hands-free rundown — listen, watch, or read the transcript.
▶ Play today's brief
New every morning, and the back catalogue is archived by date.
How outlets covered it
Meta AI Research has open-sourced Muse Glimmer, a 30-billion-parameter model designed for local execution of autonomous agents and complex tasks on consumer GPUs. This release enables developers to run advanced AI workflows directly on personal hardware, reducing reliance on cloud infrastructure and expanding the accessibility of agentic AI applications.
Meta released Muse Glimmer, a 30-billion-parameter open-weight model distilled from its larger Muse Spark model, under an Apache 2.0 license. This release highlights the legitimate use of model distillation by a single entity to create specialized, smaller models, distinguishing it from unauthorized extraction practices.
Meta announced the release of Muse Glimmer, an open-weight large language model, and committed to opening the weights for its more powerful Muse Spark 1.2 model soon. This move signals Meta's renewed focus on open-weight models, differentiating its AI strategy from companies developing proprietary models.
Meta released Muse Glimmer, a 30-billion-parameter open-weight model designed to run agentic workflows directly on local hardware. This development demonstrates a method for converting large cloud-based AI models into smaller, locally deployable agents, which could change how developers manage AI deployments and data privacy.
Meta has released Muse Glimmer, a 30-billion-parameter AI model under the Apache 2.0 license, designed to run autonomous AI agents directly on consumer hardware. This release marks Meta's return to fully open-source licensing for a major AI model, enabling local inference for sensitive agentic workflows and reducing reliance on cloud infrastructure.
Meta released Muse Glimmer, a 30-billion parameter open-weight AI model designed to power AI agents locally on consumer hardware. This model allows for privacy-sensitive, always-on personal AI agents capable of multi-step tasks without cloud reliance, aligning with Mark Zuckerberg's vision for distributed personal superintelligence.
Meta's new AI coding agent, Muse Code, is significantly cheaper than Anthropic's Fable 5, especially at its default "contributor" tier. However, the lower cost comes with the condition that user code and prompts may be used for product improvement, raising questions about data privacy and quality trade-offs.
Meta has released Muse Glimmer, a new AI model derived from its Spark 1.2 model, designed to run on a single GPU for agent-oriented tasks. This release aims to distribute AI capabilities more widely by enabling local execution on personal computers.
Meta introduced Muse Code, a new coding agent powered by Muse Spark 1.2, designed for software engineering tasks like planning, coding, and validation. The agent offers a significantly lower price point compared to competitors, but this reduced cost requires users to opt-in to share their coding data to improve Meta's models. This pricing strategy aims to compete with existing AI coding tools from companies like Anthropic and OpenAI, but raises concerns among engineering leaders regarding data privacy and security.
Meta has introduced Muse Code, an AI-powered coding agent in early beta, which uses the new Muse Spark 1.2 model for code generation and debugging. This tool aims to compete with existing coding agents by offering a more affordable pricing structure.
Meta has released Muse Code, a new AI terminal coding agent designed to assist programmers with complex software engineering tasks across large code bases. This agent, powered by Meta's Muse Spark model, aims to provide a more competitive and affordable option compared to existing AI coding tools.
Meta has released Muse Code, its first AI coding agent, developed by Meta Superintelligence Labs. This new tool aims to challenge existing AI coding assistants from companies like Anthropic and OpenAI by offering a pay-as-you-go model with a significantly cheaper contributor tier.
Meta has released Muse Code, a terminal-based AI coding agent, and Muse Spark 1.2, an updated coding-focused frontier model, entering the AI coding agent market. This move positions Meta in direct competition with existing AI coding tools from companies like Anthropic and OpenAI, marking its first significant entry into this product category.
Meta released Muse Code, a new terminal-based AI coding agent for macOS and Linux, powered by the Muse Spark 1.2 model. This tool allows developers to automate complex software engineering tasks across large code repositories, marking Meta's entry into the AI coding agent market.
Muse has released Muse Code, a beta terminal coding agent, powered by its new Muse Spark 1.2 model. Muse Code is designed to handle complex software engineering tasks across large repositories, while Muse Spark 1.2 offers improvements in code generation and debugging, making it relevant for developers and AI practitioners.