Meta Muse Glimmer 30B Hugging Face Model CardProvider report
2026-08-10Based on revision f84ecc3a0ea984a4c04542a84269e3d065350a6e of the meta-models/Muse-Glimmer-30B Hugging Face model card, captured on August 10, 2026. This snapshot includes all 24 source benchmark rows from the Benchmarks table and all 6 rows from the Preparedness chem/bio table. The paired metrics in CI Memories and Siren AgentDojo are split into separate values, producing 32 metrics and 102 scores. In the main table, Muse Glimmer uses high reasoning, while Gemma 4 31B and Qwen3.6 27B use thinking mode; comparison-model scores are retained as reported. Speed, quantization degradation, and hardware measurements are excluded because they are not model-quality benchmarks.