Meta returns to open models with Zuckerberg's plan to out-copy China and sell compute by auction
Meta just shipped a 30B agent model that runs on a MacBook. Zuckerberg's essay: we should be allowed to distill frontier labs' work. The lab-race temperature just spiked.
Why it matters
Meta re-enters the open-weight race with a compressed agent model and a public challenge to OpenAI and Anthropic's IP stance. This is both a capability milestone (consumer-hardware agent) and frontier-lab drama (distillation controversy and compute strategy).
The key facts
6 to knowMuse Glimmer: 30B agent model from Meta's Superintelligence Labs
Runs on consumer hardware; <20GB memory footprint after compression
Open-weight release strategy (Muse Spark 1.2 to follow)
Zuckerberg essay defends model distillation and calls for fewer US lab restrictions
Direct counterpunch at OpenAI and Anthropic IP/licensing positions
Compute auction strategy mentioned as part of Meta's broader plan
Go to the source
The Decoderthe-decoder.com
Publisher excerpt: Meta has released Muse Glimmer, the first open model from its new Superintelligence Labs. It's a 30B agent model that runs on consumer hardware once the weights are compressed, needing less than 20 GB of memory. In an accompanying essay, Mark Zuckerberg mounts an aggressive defense of distilling…