Meta выпустила Muse Spark 1.2 и перезапускает AI-стратегию с open-source манифестом

Meta released Muse Spark 1.2 and Muse Glimmer on August 10, 2026, and opened the weights of its flagship Muse Spark model for public access. CEO Mark Zuckerberg simultaneously published an extensive manifesto in which he acknowledged the company's lag behind competitors and proclaimed open source the only reasonable path to leadership in the AI race.

What Meta Released and How It Works

Muse Spark 1.2 is the company's most powerful open-weight model: it can be downloaded, its architecture studied, and modified without the restrictions typical of closed systems like OpenAI or Anthropic. At the same time, Meta introduced Muse Glimmer — a compact model with a permissive open-source license that runs directly on a laptop without cloud infrastructure. Glimmer was trained using distillation: a smaller model inherits knowledge from the larger Muse Spark.

Key announcement details:

• Launch date — August 10, 2026
• Muse Spark 1.2 — flagship with open weights, available for download and modification
• Muse Glimmer — lightweight version with a permissive license, runs on a personal computer
• Glimmer was created via distillation from Muse Spark
• Meta Superintelligence Labs (founded in 2025) is led by Alexander Wang, former CEO of Scale AI

Why Meta Is Falling Behind Competitors

Meta's recent models have underperformed systems from OpenAI, Anthropic, and Google on standard benchmarks, and the market has perceived Meta AI as a catch-up product. Despite the creation of Superintelligence Labs in 2025 and the hiring of Alexander Wang from his position as CEO of Scale AI, the gap with the leaders has not closed.

Pressure is also coming from the opposite direction: Chinese open models DeepSeek and Moonshot are rapidly closing in on American frontier systems. According to Anthropic, DeepSeek used the distillation method to extract characteristics from American models — the very same method Meta applies to train Muse Glimmer. In 2026, Meta's capital expenditures are estimated at $145 billion, a significant portion of which goes toward Superintelligence Labs projects.

Zuckerberg's Manifesto: Open Source as Policy

Opening the weights is only part of the announcement. Zuckerberg published an extensive text laying out both the business and political logic of the open-source approach. His central thesis: concentration of control over AI "will naturally lead to outcomes that are disadvantageous for everyone else," while most major labs are "focused on building AI for companies, governments, or other institutions."

"Our goal is for

American open-source models to be the best in the world," — Mark Zuckerberg, CEO of Meta, from the AI strategy manifesto.

The manifesto contains specific policy recommendations: protect the practice of distillation from legislative restrictions, expand energy infrastructure, and not restrict access to foreign open-source models.

"I don't think restricting access to foreign open-source models is effective.

That's how the world works, and the US won't be able to lead if it restricts itself," — Mark Zuckerberg.

What This Means

Meta is betting on openness as a strategic weapon: to attract developers, build an ecosystem around its models, and simultaneously challenge Chinese open source. Whether Muse Spark 1.2 will actually close the gap with GPT-5.5 and Claude will be shown by independent benchmarks in the coming weeks.

*Meta has been recognized as an extremist organization and is banned in Russia.