Downloads · 30 days
43
30% of all-time downloads
Muse-Glimmer-30B-FP8
dudeman2512/Muse-Glimmer-30B-FP8
Muse-Glimmer-30B-FP8 is a text generation model from dudeman2512. Use it when you need the model to write or continue text. It is set up for transformers. The card lists the license as apache-2.0.
fp8 quantization of meta-models/Muse-Glimmer-30B, produced with compressed-tensors by streaming the checkpoint tensor-by-tensor (the model is never fully instantiated).
All-time downloads
143
Public
Parameters
29.8B
34.4 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors34.4 GB · 100%
Parameter types
How the weights are stored.
F8_E4M325.2B · 85%
Try a prompt
Base models
Model card
From the Hugging Face model README
Muse-Glimmer-30B-FP8
fp8 quantization of meta-models/Muse-Glimmer-30B, produced with compressed-tensors by streaming the checkpoint tensor-by-tensor (the model is never fully instantiated).
8-bit float weights (float8_e4m3fn), per output channel, with dynamic per-token activation quantization. Highest fidelity of the set and the largest.
All quantizations of this model
| Variant | Format | Size | vs BF16 | Mean rel. error | Linears quantized | Left BF16 |
|---|---|---|---|---|---|---|
| Muse-Glimmer-30B-FP8 ← this one | float-quantized | 34.40 GB | 58% | 0.0266 | 416 | 0 |
| Muse-Glimmer-30B-NVFP4 | nvfp4-pack-quantized | 23.38 GB | 39% | 0.0947 | 416 | 0 |
| Muse-Glimmer-30B-INT4-W4A16 | pack-quantized | 22.20 GB | 37% | 0.1156 | 416 | 0 |
Mean relative error is ||dequant(W) - W|| / ||W||, averaged over a sample of quantized Linear layers, measured against the original BF16 weights. Lower is better.
This variant
| Format | float-quantized |
| Weight bits | 8 |
| Group size | per-channel (no grouping) |
| Strategy | channel |
| Linears quantized | 416 |
| Left in BF16 | 0 |
| Shards | 8 |
| On disk | 34.40 GB |
| Mean relative error | 0.0266 |
| Shape/dtype conformance failures | 0 |
Use with vLLM
vllm serve dudeman2512/Muse-Glimmer-30B-FP8
How this was made
Every produced tensor is checked for shape/dtype conformance against what the server expects, then reconstruction error is measured against the source BF16 weights, before anything is published. The numbers in the table above are those measurements — not estimates.
- Task
- Text Generation
- Library
- transformers
- Type
- muse_glimmer
- License
- apache-2.0
- Created
- Aug 22, 2026
- Updated
- Aug 22, 2026