Meta · Model release
Muse Glimmer 30B: Apache-2.0 local agent weights
Muse Glimmer 30B is an open-weight multimodal model described by its release materials as a Muse Spark distill. Hardware fit depends on quantisation, runtime and context; Meta publishes no hosted per-token list for Glimmer.
What changed
- Apache 2.0 weights announced 2026-08-10.
- This is not Muse Spark API pricing — hosted inference is billed by the serving provider.
- Quantized formats can reduce memory needs, but the required hardware depends on the exact checkpoint, quantization, runtime and workload.
Best fit
- Local agents
- On-device coding + tool use
- Privacy-sensitive multimodal work
What to be careful about
- Do not treat one GPU-memory estimate as a guaranteed fit; benchmark the exact published artifact and runtime.
- Agentic quality is not Muse Spark API quality.
Dataset snapshot: Muse Glimmer 30B is also represented in the checked comparison dataset. Prices and benchmark figures carry their own field dates.