Observed Signal · Oct 15, 2026 · Partnership · Source: CNBC Technology · Impact: 4/5 · Sentiment: Positive
Apple and Meta partner to run Muse AI on Macs
Apple and Meta have announced a strategic collaboration to run Meta's open-weight AI model, Muse Spark Open, locally on Apple's Mac Studio computers, signaling a thaw in their historically tense relationship. This 816-billion-parameter, natively multimodal model features a 1 million token context window, 42 billion active parameters, and is optimized for complex coding and agentic workflows, reportedly outperforming other open-weight models but trailing Chinese models like MiMo-V2.6-Pro. To run at full capacity, the model requires two Mac Studio systems with M5 Ultra chips and 512 GB of unified memory each (1 TB total), but Apple's 512 GB configuration is not yet available, expected at the end of October. Current Mac Studio options max out at 256 GB. Configurations start at $10,799, and the model is also compatible with AMD, Intel, and NVIDIA chips. Available via Hugging Face, Ollama, and cloud platforms, it is restricted by the Muse Model License for large-scale providers.
This partnership between two tech giants signals a major shift in AI alliances and highlights Apple's growing role in AI hardware, potentially challenging Nvidia's dominance. It also shows Meta's continued push for open-weight AI models.
Track Meta Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Apple and Meta collaborated to run Meta's open-weight AI model, Muse Spark Open, on Mac Studio computers (announced October 15, 2026).
- Muse Spark Open has 816 billion parameters, 42 billion active parameters, a 1 million token context window, and is natively multimodal for coding and agentic tasks.
- The model runs at full capacity on two Mac Studio systems with M5 Ultra and 512 GB RAM each (1 TB total), but Apple's 512 GB configuration is not yet available until end of October.
- The model is available on Hugging Face, Ollama, and major cloud platforms, with optimizations for AMD, Intel, and NVIDIA chips.
- The Muse Model License restricts Model-as-a-Service use for companies with over $20 million in annual revenue, and the model trails Chinese models like MiMo-V2.6-Pro on benchmarks.
Connected Companies & Entities
22 Entities mapped“Meta and Apple revealed a new collaboration Thursday that will enable developers to easily run Meta's new open-weight AI model on Apple's hi...”
“Apple and Meta revealed a new collaboration Thursday that will enable developers to easily run Meta's new open-weight AI model on Apple's hi...”
“who joined Meta when the company invested more than $14 billion into his former startup Scale AI last year.”
“Apple sued OpenAI in July, alleging trade secret theft, a claim OpenAI denied.”
“It could also signal Apple's growing strength in AI chips, a market that Nvidia has dominated so far due to their prevalence in large data c...”
“Earlier this year, Apple announced a deal with Google, to use its Gemini models in Apple Intelligence as well as Google Cloud infrastructure...”
“Thinking Machines' Inkling model is mentioned in benchmark comparisons.”
“Meta offers the model with optimizations for AMD chips.”
“The model can be downloaded via tools like Ollama.”
“Meta offers the model with optimizations for Intel chips.”
“Xiaomi's MiMo-V2.6-Pro is listed as the top open-weight model in the Artificial Analysis ranking.”
“The model can be downloaded from Hugging Face.”
“DeepSeek V4.1 Flash is listed in the Artificial Analysis ranking.”
“Alibaba's Qwen model is mentioned as having a more permissive license.”
“Mistral Large 4 is compared with Muse Spark Open in benchmarks.”
“Moonshot AI's Kimi K3 model is mentioned in benchmark comparisons.”
Ontology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Meta launches Muse Spark AI model
Meta released Muse Spark, the first AI model from its new Superintelligence Labs led by Alexandr Wang, following a multibillion investment in Scale AI that secured roughly half the company. Muse Spark will be integrated into Meta AI and across Meta apps (Instagram, Facebook, WhatsApp, Threads) to provide faster, more personalized and visual responses, including the ability to embed Reels, photos and posts in replies while crediting creators. Meta says the model can support health-related answers after training with more than 1,000 physicians, and includes a Shopping mode plus partial agentic task capabilities. Meta positions Muse Spark as an efficiency-focused step rather than a competitor-matching release; the model is proprietary (a departure from the open-source Llama approach), though future versions could return to open source.
Meta open-sources Muse Spark, launches Muse Glimmer
Muse Glimmer is a roughly 30‑billion‑parameter dense multimodal (text+image) model released by Meta under an Apache‑2.0 license with open weights and single‑consumer‑GPU runtimes for Mac/PC. Engineered as a local "brain" for agentic, always‑on assistant workflows, it supports multi‑step reasoning, tool access, task diagnosis/retry, very large 128K context windows and was trained in 100+ languages. Glimmer was distilled from a larger closed Muse Spark and uses aggressive 4‑bit quantization plus a DFlash speculative‑decoding drafter to reach an on‑device footprint near ~20 GB (BF16 ~60 GB, 4‑bit ~18 GB). Meta reports >57 tokens/sec in internal tests and claims wins versus Gemma4‑31B and Qwen3.6‑27B; the model is distributed via Hugging Face and multiple runtimes/partners. Its agentic, deep‑access design raises substantial security and privacy engineering requirements. Reported by HORIZONT (2026‑08‑17), Denise Samer.
Meta Expands Muse AI Agent with Avatars, Mac Control, and Commerce
At its Connect event, Meta unveiled major updates to its personal AI agent, Muse, including a new real-time avatar feature, macOS integration, and a dedicated email address. The company also announced partnerships with Stripe, Shopify, PayPal, and retailers like Best Buy, Walmart, and Sephora to enable agentic shopping. Meta aims to make Muse a central platform for everyday tasks, leveraging its Muse Spark multimodal model. CEO Mark Zuckerberg emphasized a long-term vision of Muse as a 'personal superintelligence,' with monetization expected through transaction fees. The announcement also highlighted the opening of its connector platform to developers, which has already received over 1,500 applications.
Receive all future signals for Meta & competitors in your briefing
Never miss material M&A deals, leadership changes, account wins, or product rollouts. Polaris7 monitors 20,000+ primary sources 24/7, delivering verified telemetry directly into your executive briefing.