Meta launches Muse Spark Open, claiming it as the most powerful open-weight AI model in the U.S.

icon币界网
Share
This is the logo of the coin.
META
----
AI summary iconSummary
Meta unveiled Muse Spark Open, claiming it as the most powerful open-weight AI model in the U.S., featuring 816 billion parameters and a 1 million token context window. The model supports code generation and agent collaboration, and can run locally on two Apple M5 Ultra chips with 512GB of memory via the MLX framework. Developers can access it through API platforms and download weights from Amazon SageMaker and Hugging Face. It is compatible with AMD, Apple, Intel, and NVIDIA chips. Meta enhanced safety alignment, achieving strong results on benchmark tests. The model is available under the Muse Model License, largely free for developers. This AI + crypto news update highlights recent on-chain developments.
CoinDesk reports:

IT Home, October 16: On October 15 local time, Meta released the open-weight AI model Muse Spark Open. Meta officially claims that Muse Spark Open is the most powerful open-weight model from the United States to date.

Muse Spark Open employs a native multimodal architecture with a total of 816B parameters, including 42B activated parameters, and supports a context window of 1M tokens. Designed specifically for code generation and agent collaboration, the model demonstrates exceptional performance in handling long-range tasks, complex software engineering, and multi-step reasoning.

Thanks to the latest MLX framework, Muse Spark Open is the first model in its class to run locally on desktop using just two Apple M5 Ultra chips and a 512 GB RAM Mac Studio.

Starting today, developers can access Muse Spark Open via API platforms including Baseten, Fireworks AI, Groq, Modal, Nebius, OpenRouter, and Together AI; model weights are also now available for public download on Amazon SageMaker AI, Google Cloud Model Garden, Hugging Face, and Microsoft Foundry Models.

The model has been optimized for leading inference engines including SGLang, vLLM, TensorRT-LLM, Hugging Face Transformers, Apple MLX, llama.cpp, and Unsloth, and has undergone deep optimization for AMD, Apple, Intel, and NVIDIA chips. Developers can also easily perform secondary development and customization using the PyTorch framework.

Muse Spark Open underwent targeted safety alignment during the post-training phase. Benchmark results demonstrate that the model provides robust safety protections across all tested risk dimensions.

The model weights for this release are made available under the Muse Model License, which includes restricted commercial terms, and are free for the vast majority of developers to access and use.

IT Home provides the following related test data:

Source:Show original
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.