Meta ships open model Glimmer as Zuckerberg pitches 'AI for everyone'
Meta released the open-weight model Glimmer this week, letting anyone download and run it on their own hardware, in contrast to Muse Spark, the company's more powerful model that stays locked behind its own APIs. The release landed alongside a roughly 6,500-word letter from Mark Zuckerberg arguing that AI should be "for everyone" rather than controlled by a handful of labs. On TechCrunch's Equity podcast, hosts Kirsten Korosec, Anthony Ha, and Rebecca Bellan discussed Glimmer and Zuckerberg's manifesto alongside the week's other headlines. They said Zuckerberg's "for everyone" framing comes with some asterisks, though the episode description does not spell out what those caveats are. The same episode also covered the true cost of the AI industry's energy needs and a $250M acquisition the hosts described as having gone very wrong; no figures are given for the energy costs, and no acquirer or target is named for the deal.
Key facts
- Meta released Glimmer, an open-weight AI model anyone can download and run on their own hardware, as a contrast to Muse Spark, Meta's more powerful model that stays locked behind its own APIs.
- Mark Zuckerberg published a roughly 6,500-word letter alongside the release arguing AI should be "for everyone" rather than controlled by a handful of labs.
- TechCrunch's Equity podcast, hosted by Kirsten Korosec, Anthony Ha, and Rebecca Bellan, discussed the release and said Zuckerberg's vision "comes with some asterisks."
- The same episode covered the true cost of the AI industry's energy needs and a $250M acquisition the hosts called gone very wrong, without naming figures, an acquirer or a target.
Why it matters
Glimmer gives Meta an open-weight release it can point to as evidence of Zuckerberg's stated position that AI access should not sit with a handful of labs, even as the company keeps its stronger model, Muse Spark, closed behind APIs. The pairing of a free download with a lengthy personal manifesto turns a model release into a statement about how Meta wants to be seen in the wider argument over open versus closed AI.
Who it affects
Developers and researchers who want to download and run a Meta model on their own hardware rather than through a locked API gain a new open option in Glimmer. The framing also matters to anyone tracking how the large labs justify keeping their most capable systems closed while promoting openness elsewhere.
How to use it
The source describes Glimmer as downloadable and runnable on a user's own hardware, unlike Muse Spark, which stays behind Meta's APIs. No technical detail, such as parameter count, architecture or benchmark results, is given for either model, and no licence terms or pricing are mentioned.
How solid is it
The material comes from TechCrunch's Equity podcast, a named editorial team (Kirsten Korosec, Anthony Ha, Rebecca Bellan) discussing the release as part of a weekly news roundup, not a first-hand technical writeup. No release date beyond "this week" is given, and several details discussed on the episode, including the specifics of the acquisition and the energy-cost figures, are not spelled out in the available text.
Risks and caveats
The hosts flag that Zuckerberg's "for everyone" vision "comes with some asterisks," but the source does not say what those caveats specifically are, so the substance of that skepticism is not verifiable from this text alone. Readers should note that Meta's most capable model remains closed even as the company promotes openness through Glimmer.
“the vision comes with some asterisks”
— TechCrunch's Equity podcast hosts