Meta has introduced the newest AI model: Muse Glimmer, developed by its Meta Superintelligence Labs. It is designed to handle functions including agentic task completion, multi-step reasoning, and multi-model input and reasoning.
Announcement of Muse Glimmer
On Monday, Meta announced Muse Glimmer. The latest AI model from Meta Superintelligence Labs has the model weights under a permissive Apache 2.0 license. Muse Glimmer is a 30 billion-parameter model that can run on a Mac or PC with a single consumer GPU.
Users can download and customise the Glimmer, a distilled version of the company's Muse Spark 1.2 model designed with a focus on efficiency to minimise system requirements.
Mark Zuckerberg, Meta Chief Executive Officer, penned 6500 words ahead of the launch of the Muse Glimmer AI model, which challenges a vision of AI championed by startups like OpenAI and Anthropic PBC, which largely favour keeping the technology under tight control.
Google offers its Gemma family of open-weight models that can be run on local devices. In contrast, low-cost open models from Chinese companies including DeepSeek, Moonshot and Alibaba Group Holding Ltd. have intensified pressure on US rivals who charge for access.
Features and components of Glimmer Muse
The Meta model would require more than 55 GB of memory at full precision. Meta uses 4-bit quantisation to reduce the model size to under 20GB, and this lets it run on a 24GB or 32GB memory setup.
Meta has also shared a direct comparison of the Muse Glimmer with Google's Gemma4-31B and Alibaba's Qwen3.6-27B. The results show that Muse Glimmer dominates both models in 12 tests.
According to Meta, Muse Glimmer is built for end-to-end agentic task completion, reliable tool use, multi-model reasoning, error diagnosis and tool retries, multi-step reasoning, among others. It is also trained on data from over 100 languages.
Users can download the weights from Hugging Face. Meta has announced that the latest Muse Glimmer will be available in the coming days from apps like Ollama, LM Studio, and Unsloth.
The tech giant, Meta, confirmed that it is working with AMD, Arm, Dell, Intel, and Nvidia to optimise performance across devices. Meta confirmed that the optimised integrations for llama. MLX, and ExecuTorch will be available in the coming days.
Key Features of Muse Glimmer
End-to-end task completion: Handles complex tasks from start to finish.
Strong benchmarks: Performs well on DeepSearch QA, MCP-Atlas, T-Bench and SWE-Bench.
Reliable tool use: Executes different function calls using precise tool schemas.
Long-horizon reasoning: Maintains plans across extended, multi-step workflows.
Error recovery: Diagnoses failed tool calls and retries instead of stopping.
Visual understanding: Processes text, screenshots, charts and documents together.
Agentic workflows: Supports OpenClaw and other orchestration patterns.
Flexible reasoning: Offers different reasoning strengths for speed or quality.
Multilingual: Trained on data covering more than 100 languages.
Future Plans of Meta
Meta is also developing plans for a cloud infrastructure business that will sell access to AI computing power and models, along the lines of Amazon Web Services.
However, the lack of detail on the plan from Zuckerberg has frustrated investors who worry about the company achieving a return on investment.
The company said on Monday that it intends to set up a $1 billion fund to invest in US communities where it owns and operates data centres.
Major infrastructure projects proposed by companies like Meta and Google have been met with resistance in many locations across the US as detractors protest against data centres' use of land, energy and water.
