Muse Glimmer 30B

Meta AI🇺🇸 United States
active
Context window131K tokens
Input / 1M tokens$0.35
Output / 1M tokens$1.5

Version History

30Bmajor

New 29.6B-parameter open-weight model with a built-in perception encoder, distilled from Muse Spark, designed for on-device agentic use with 4-bit quantization and speculative decoding.

Benchmark Scores

Full leaderboard →
94.7%
AIME 2026
83.5%
GPQA
74.0%
MMMU
43.2 tokens_per_sec
Speed (tok/s)
76.0%
SWE-bench Verified

Coverage

model release

Meta Releases Muse Glimmer 30B, an On-Device Agentic Model with Built-In Perception Encoder

Meta Superintelligence Lab has released Muse Glimmer, a 29.6-billion-parameter multimodal model distilled from Muse Spark for autonomous agentic tasks that run entirely on consumer hardware. The Apache 2.0-licensed model ships with a dedicated perception encoder, 131K+ token context, and speculative decoding for local speedups up to 3.1x.

3 min read