Apple-Meta collaboration announced to run Muse Spark Open locally on Mac Studios, described as the first public cooperation between the two companies since 2017.
Following·Day 191·first covered 8 Apr 2026·5 sources
Meta Releases Muse Spark Open, Claims Most Capable US Open-Weight Model
The gist
Meta's release makes a large multimodal mixture-of-experts model freely available as open weights, which expands access to frontier-class AI capabilities outside of closed APIs. The Apple partnership extends local inference to high-end consumer hardware, broadening deployment options beyond cloud infrastructure.
The full picture
Meta released Muse Spark Open on October 15, 2026, describing it as 'the most capable US open-weight model to date' and built for coding and agentic work. The model uses a mixture-of-experts architecture with 816 billion total parameters and 42 billion active parameters, supports a 1 million token context window, and handles text, image, video, and audio natively. Apple and Meta are collaborating to allow the model to run locally on 2 to 4 connected Mac Studios, described as the first public cooperation between the two companies since 2017. Benchmark evaluations of earlier Muse Spark models show mixed results compared to competitors: Muse Spark scored third on HLE reasoning at 39.9%, behind Gemini 3.1 Pro Preview (44.7%) and GPT-5.4 (41.6%), and trailed on agentic benchmarks. Muse Spark 1.1 led the health category on HealthBench Professional but ranked below Claude Opus 4.8 and GPT-5.5 on several multimodal and reasoning evaluations.
How it developed
Muse Spark 1.1 benchmark results published, showing mixed performance versus Claude Opus 4.8 and GPT-5.5 with a lead in the health category.
Artificial Analysis publishes benchmark evaluation of Muse Spark, placing it third on HLE and noting weak agentic performance.
Sources
HuggingFace Blog
Meta AI Twitter
Milk Road AI Twitter
artificialanalysis.ai
The daily email
Want this in your inbox?
I send one email each morning with the stories that moved. If you would rather just read here, that works too.
Subscribe free