Muse Glimmer 30B GGUF
Certified GGUF precision quantization ladder for meta-models/Muse-Glimmer-30B dense multimodal architecture.
Certified GGUF precision variants (Q8_0, Q4_K_M, Q2_K) of meta-models/Muse-Glimmer-30B evaluated across standard reasoning, mathematics, and science benchmarks (ARC-MC, GSM8K, MATH-500) under reproducible seed controls.
Architecture
- Parameters
- 29.60B
- Active
- 29.60B (100%)
- Layers
- 52
- Quantization ladder
- Certified GGUF
Downloads
Exact release files
3 immutable artifacts
muse-glimmer-30b.q4_k_m.gguf
Size: 17 GB
20–24 GB reference capacity. Balanced agentic performance and multimodal retention.
SHA-256 Checksum
6d8112c565f5c73304b05b64aff22e8197aa7c37a25436d25846e19793a4876a
muse-glimmer-30b.q8_0.gguf
Size: 30 GB
32–48 GB reference capacity. Near-lossless precision for production and critical reasoning.
SHA-256 Checksum
9c8ed26257c93a201ce6db14a4467d926439fe68e3f0cbed1b06ea61af193a00
muse-glimmer-30b.q2_k.gguf
Size: 11 GB
12–16 GB reference capacity. Memory-constrained consumer workstations.
SHA-256 Checksum
2e294ba3c5db8e97c91ca3b381f2859be0d2ab309565970ebe05b89466e7a191
Deployment
Quickstart & CLI
llama.cpp & Ollama
Stream weights directly from Hugging Face via Ollama CLI:
ollama run hf.co/Omega-Factory/muse-glimmer-30b-GGUF:Q4_K_MRun inference locally using llama.cpp with the recommended artifact:
llama-cli -hf Omega-Factory/muse-glimmer-30b-GGUF -m muse-glimmer-30b.q4_k_m.gguf -p "Explain the core mechanics of mixture-of-experts models." -n 512Launch a local OpenAI-compatible HTTP server:
llama-server -hf Omega-Factory/muse-glimmer-30b-GGUF -m muse-glimmer-30b.q4_k_m.gguf --port 8080 -ngl 99Evidence
Measured outcomes & precision ladder
Precision & Accuracy Matrix vs. FP16 Baseline
Holdout evaluation split (N=200 per category, seed 42) comparing all quant variants to unquantized baseline.
ARC-MC
95.5%
GSM8K
95.5%
MATH-500
65.0%
ARC-MC
95.5%
GSM8K
95.5%
MATH-500
64.0%
-1.0 pp
ARC-MC
96.0%
+0.5 pp
GSM8K
93.0%
-2.5 pp
MATH-500
62.0%
-3.0 pp
ARC-MC
92.0%
-3.5 pp
GSM8K
74.0%
-21.5 pp
MATH-500
32.0%
-33.0 pp
Item-Level McNemar Discordance & Flips vs. FP16 Baseline
Per-item analysis measuring individual test questions that flipped answer status relative to the unquantized FP16 baseline (N=200).
Recommended Release Rung Scoped Verification (Q4_K_M)
Formal cryptographic release marker evaluation records with 95% confidence intervals and sample counts.
ARC-Challenge Science Reasoning
muse-glimmer-30b-arc-mc-v1 · accuracy_pct
Base
95.50%
Release
96.00%
Delta
+0.50 pp
Holdout evaluation set (N=200, seed 42) comparing Q4_K_M quantized release against uncompressed FP16 baseline under standard multiple-choice log-likelihood scoring.
95% CI [-3.28, 4.28] pp · 200 samples · 1 seeds · artifact 6d8112c565f5c73304b05b64aff22e8197aa7c37a25436d25846e19793a4876a
GSM8K Grade School Math
muse-glimmer-30b-gsm8k-v1 · accuracy_pct
Base
95.50%
Release
93.00%
Delta
-2.50 pp
Holdout evaluation set (N=200, seed 42) comparing Q4_K_M quantized release against uncompressed FP16 baseline under greedy decoding.
95% CI [-6.95, 1.95] pp · 200 samples · 1 seeds · artifact 6d8112c565f5c73304b05b64aff22e8197aa7c37a25436d25846e19793a4876a
MATH-500 Advanced Mathematics
muse-glimmer-30b-math500-v1 · accuracy_pct
Base
65.00%
Release
62.00%
Delta
-3.00 pp
Holdout evaluation set (N=200, seed 42) comparing Q4_K_M quantized release against uncompressed FP16 baseline under greedy decoding.
95% CI [-12.35, 6.35] pp · 200 samples · 1 seeds · artifact 6d8112c565f5c73304b05b64aff22e8197aa7c37a25436d25846e19793a4876a
Literature & Architecture
Foundational Research & Citations
Primary scientific publications & architecture preprints
DFlash: Block Diffusion for Flash Speculative Decoding
Jian Chen, Yesheng Liang, Zhijian Liu
Speculative decoding framework using lightweight block diffusion for parallel token drafting, achieving over 6x lossless acceleration and integrated into Muse Glimmer's native speculative architecture.
Perception Encoder: The best visual embeddings are not at the output of the network
Daniel Bolya, Po-Yao Huang, Peize Sun, et al.
State-of-the-art vision perception encoder (ViT-G/14, ~1.8B parameters) powering Muse Glimmer's multimodal interleaved image and text understanding.
Which Quantization Should I Use? A Unified Evaluation of llama.cpp Quantization
Empirical GGUF Research Study
Comprehensive empirical analysis of llama.cpp GGUF quantization schemes (Q8_0, Q4_K_M, Q2_K) across reasoning benchmarks, memory footprints, and CPU/GPU throughput.
Limitations
- MULTIMODAL_PERCEPTION_ENCODER — Requires ViT-G/14 perception encoder integration for multimodal image input processing.
- VRAM_ESTIMATE_NOTICE — Artifact size is disk usage, not a measured VRAM requirement. Deployment requires peak-memory profiling.
- COMPRESSED_DERIVATIVE — Pruned and re-quantized derivative of meta-models/Muse-Glimmer-30B under Apache 2.0 license.
Public release identity
- Release ID
- muse-glimmer-30b-gguf-v1
- Release version
- 1
- Published
- 2026-10-12T07:13:56.000Z
- Model ID
- Omega-Factory/muse-glimmer-30b-GGUF
- Base model ID
- meta-models/Muse-Glimmer-30B
- Catalog family
- Not supplied
- Quantization rung
- Not supplied
- Supersedes
- Not supplied
- HF repository
- Omega-Factory/muse-glimmer-30b-GGUF
- HF repository URL
- https://huggingface.co/Omega-Factory/muse-glimmer-30b-GGUF
- HF revision
- 8f3d41daea566b7399b93e8cdd16be427b7e6c07
- HF revision URL
- https://huggingface.co/Omega-Factory/muse-glimmer-30b-GGUF/tree/8f3d41daea566b7399b93e8cdd16be427b7e6c07
- Release manifest SHA-256
- 413ff55981e9ed63e0bde2f10e4e96d269aea4fbba2120db82eb9328224b7a12
- Catalog SHA-256
- b7dde414fe014d98482de94c7086aa12303312eed35bcb2132272d76dea8141b
- Catalog generated
- 2026-10-11T12:45:00Z
- License
- Apache 2.0
- License ID
- apache-2.0
- Commercial use
- ALLOWED
- License URL
- https://www.apache.org/licenses/LICENSE-2.0