Meta Releases Muse Glimmer, a 30B Open-Weight Model for Consumer GPUs

Muse Glimmer is a 30B dense model designed for local, multi-step agentic tasks that fits on a single consumer GPU. Zuckerberg positioned it as a return to open weights, promising that Muse Spark 1.2 — Meta's newest foundation model — will also be opened soon. The release landed on SageMaker JumpStart the same week (as Muse-Glimmer-30B) and was accompanied by Meta shipping muse image to its model API at $0.01/image, one of the best price-to-quality ratios for production volume, per Alexandr Wang.

The developer community embraced the dense, single-GPU design, which sidesteps the memory overhead of large MoE deployments and suits on-device and privacy-sensitive workloads. It reignited the local-AI conversation that has run hot all week alongside 'Small Models Have Arrived' (766 pts on HN) and GLM-5.3-Flash benchmarks.

Competitively, the move is notable given Meta's earlier retreat from open releases and its enormous capex — guidance raised to $130–$145B for 2026, funded partly by monetization pushes like new Meta AI subscription tiers. Muse Glimmer also positions Meta against Alibaba's Qwen and Zhipu's GLM in the open-weight race. The caveat: Meta has promised open weights before and delayed; the real test is whether Muse Spark 1.2's weights actually ship on the timeline Zuckerberg implied.