August 10, 2026
Open-source or open season?
Meta Muse Glimmer – open weights 30B local coding model
Meta drops a new free AI brain, and the crowd is already arguing if it’s a hero or just hype
TLDR: Meta released Muse Glimmer, a new free-to-use AI model meant to run on personal computers instead of the cloud. Commenters are split between celebrating Meta’s open-source return, questioning whether the performance numbers are overhyped, and arguing that 30B-size models may still be too small to truly impress.
Meta just tossed a fresh AI model called Muse Glimmer into the wild, and the real show instantly moved to the comments. On paper, this is a 30-billion-parameter coding-focused model that Meta says can run locally on regular high-end personal hardware, not just in giant corporate data centers. In plain English: the company is pitching an AI helper you can use on your own machine, even without internet, and it’s giving away the model weights under an Apache 2.0 license so developers can tinker freely.
But the community didn’t just clap politely — it went straight into open-source soap opera mode. One camp was thrilled, basically cheering, “Meta did not abandon open source,” with several commenters treating this like a reassuring comeback for free and open AI releases. Another group immediately turned into skeptical sports commentators, side-eyeing the benchmark scores and wondering whether the results are too polished to trust. One comment flat-out hinted it might be “benchmaxed,” which is tech-for-“did they train for the test?”
Then came the rivalry energy. People are already lining Muse Glimmer up for a cage match with Qwen’s next release, while others are reviving a familiar complaint: 30B is nice, but is it actually smart enough? One commenter basically said the size sounds convenient, but still not powerful enough for serious everyday use. The funniest part? Even the optimists sound competitive, like they’re watching AI brands the way sports fans watch transfer season. In other words: Meta released a model, and the internet released opinions.
Key Points
- •Meta released Muse Glimmer, a 30B-parameter open-weight model, under an Apache 2.0 license.
- •The model is designed for always-on local agent workflows and is intended to run on a Mac or PC with a single consumer GPU.
- •Meta says Muse Glimmer supports local agents, function calling, local coding, LLM-as-a-judge evaluation, multimodal input, and long-horizon reasoning.
- •Training included pre-training on Muse Spark outputs via logit distillation, mid-training on longer-context agent-heavy data, and post-training with supervised fine-tuning, on-policy distillation, and reinforcement learning.
- •Meta says the model was evaluated using its Advanced AI Scaling Framework and released on Hugging Face with planned integrations for llama.cpp, MLX, and ExecuTorch.