Вход на сайт

Просмотр новости

Найдите то, что Вас интересует

Meta unveils Muse Spark Open to compete with Chinese open weights

Дата публикации: 16-10-2026 07:35:15

The multimodal model has 816bn total parameters, with 42bn active parameters and a 1m-token context window.
Read more: Meta unveils Muse Spark Open to compete with Chinese open weights


Основное содержимое страницы с новостью.

The multimodal model has 816bn total parameters, with 42bn active parameters and a 1m-token context window.

Meta has unveiled Muse Spark Open, calling it the “most capable” US open-weight model to date. The company said that the new model is the first in its class to be able to run on a desktop computer using two of Apple’s M5 Ultra Mac Studios with its latest machine learning framework.

The multimodal model has 816bn total parameters, with 42bn active parameters and a 1m-token context window. According to the company, it is built for coding, agentic work and other complex, multi-step reasoning tasks.

Unlike their closed-weight contemporaries, developers can download parameters of an open-weight model and modify them for their needs. Open-weight models, however, are not the same as open-source models that also provide access to training data and code.

Meta said that the model weights are available for download via Amazon SageMaker AI, Google Cloud Model Garden, Hugging Face and Microsoft Foundry Models. The model is available through popular API platforms including Groq, OpenRouter and Baseten.

China currently leads the open-weights space, with models developed by Chinese makers such as Xiaomi, Z AI, Kimi, Alibaba and DeepSeek routinely ranking above their US counterparts in Artificial Analysis’ ranking.

Muse Spark Open has had additional post-training safety, the company said. It evaluated the model across its “most severe risks” and found that it met the thresholds set by Meta.

Open-weight models are considered necessary for AI leadership, as being able to tweak a model means organisations, from start-ups to public institutions, can build on advanced models without training one from scratch.

Earlier this month, France’s AI darling Mistral announced its latest open-weight AI model Mistral Large 4, which it said “significantly outperforms” any open-weight model developed in the US or Europe.

Known as ‘Le Chonk’, ML4 is a 1trn-parameter natively multimodal model with 49bn active parameters, and Mistral’s “largest and most capable model to date”.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.

Схожие новости

#Наименование новостиТональностьИнформативностьДата публикации
1Meta бросила вызов китайским ИИ-моделям с новой Muse Spark Open014.9116-10-2026
2Meta releases open-weight Muse Spark model for coding, agentic tasks 014.9616-10-2026
3Meta draait AI-model met 816 miljard parameters op twee Macs012.5316-10-2026
4Meta enters AI for developers race with new Muse Spark model0709-07-2026
5Meta Publishes Muse Glimmer As 30B Open Agentic Model013.9510-08-2026
6Meta* выпустила Muse Glimmer — открытую версию своей лучшей ИИ-модели08.6311-08-2026
7Meta Releases Muse Glimmer, a 30-Billion-Parameter Open-Weight AI Model That Runs on a Single Consumer GPU05.0611-08-2026
8Meta's new Muse Spark Open AI needs two $5,499 Mac Studios to run on desktop019.9915-10-2026
9Meta dévoile une IA "open-weight" plus légère010.8311-08-2026

Классификация: Пресс-релизы. Схожих патентов: 0. Схожих новостей: 9. Тональность: 0. Информативность: 6.41. Источник: www.siliconrepublic.com.