Instructions to use Dyluhn/Muse-Glimmer-30B-R9V-V1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use Dyluhn/Muse-Glimmer-30B-R9V-V1 with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Dyluhn/Muse-Glimmer-30B-R9V-V1
# Run inference directly in the terminal:
llama cli -hf Dyluhn/Muse-Glimmer-30B-R9V-V1

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Dyluhn/Muse-Glimmer-30B-R9V-V1
# Run inference directly in the terminal:
llama cli -hf Dyluhn/Muse-Glimmer-30B-R9V-V1

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf Dyluhn/Muse-Glimmer-30B-R9V-V1
# Run inference directly in the terminal:
./llama-cli -hf Dyluhn/Muse-Glimmer-30B-R9V-V1

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf Dyluhn/Muse-Glimmer-30B-R9V-V1
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Dyluhn/Muse-Glimmer-30B-R9V-V1

Use Docker

docker model run hf.co/Dyluhn/Muse-Glimmer-30B-R9V-V1

LM Studio
Jan

vLLM

How to use Dyluhn/Muse-Glimmer-30B-R9V-V1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Dyluhn/Muse-Glimmer-30B-R9V-V1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Dyluhn/Muse-Glimmer-30B-R9V-V1",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/Dyluhn/Muse-Glimmer-30B-R9V-V1

Ollama
How to use Dyluhn/Muse-Glimmer-30B-R9V-V1 with Ollama:
```
ollama run hf.co/Dyluhn/Muse-Glimmer-30B-R9V-V1
```
Unsloth Desktop

How to use Dyluhn/Muse-Glimmer-30B-R9V-V1 with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Dyluhn/Muse-Glimmer-30B-R9V-V1

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "Dyluhn/Muse-Glimmer-30B-R9V-V1"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use Dyluhn/Muse-Glimmer-30B-R9V-V1 with Docker Model Runner:
```
docker model run hf.co/Dyluhn/Muse-Glimmer-30B-R9V-V1
```

Lemonade

How to use Dyluhn/Muse-Glimmer-30B-R9V-V1 with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Dyluhn/Muse-Glimmer-30B-R9V-V1

Run and chat with the model

lemonade run user.Muse-Glimmer-30B-R9V-V1-{{QUANT_TAG}}

List all available models

lemonade list

Hermes Agent

How to use Dyluhn/Muse-Glimmer-30B-R9V-V1 with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Dyluhn/Muse-Glimmer-30B-R9V-V1

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default Dyluhn/Muse-Glimmer-30B-R9V-V1

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use Dyluhn/Muse-Glimmer-30B-R9V-V1 with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Dyluhn/Muse-Glimmer-30B-R9V-V1

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "Dyluhn/Muse-Glimmer-30B-R9V-V1" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Muse Glimmer 30B R9V V1

Rough-draft release. This is the first public R9V model package, not the best available Muse Glimmer quant. It currently loses badly to Unsloth's higher-fidelity Q5/Q6 work in the quality evaluation used here. Do not infer quality superiority from the R9V name or the V1 label.

V1 is the public release name. V12 is the internal research lineage of the exact GGUF and remains in manifests and hashes for reproducibility.

The complete package, including the optional projector and DFlash sidecar, is pinned at Hugging Face revision 093f8ced7a8e2308b0f597084ebdbfa5f6614f75.

What it is

24,554,611,392-byte GGUF
SHA256 f4870ff4ac316c1dbf50a55501f4c00e16070336fc40e119ff1167e43382856a
731 tensors: 313 F32, 351 Q8_0, and 67 Q4_K
seven gate/up pairs promoted to Q8_0 at layers 11, 15, 19, 23, 27, 31, and 35
derived from Unsloth's pinned Q8_0 Muse GGUF with llama.cpp at the pinned revision in sources.lock.json

Quality disclosure

All figures below use the same native-BF16 teacher, evaluator, 480 chunks, 512 tokens per chunk, and 122,400 evaluated positions.

Quant	Bytes	Mean KLD (lower is better)	Same top prediction
R9V V1 / V12	24,554,611,392	0.006121	96.879%
Unsloth Q4 comparison	15,878,222,368	0.016883	94.806%
Unsloth UD-Q5_K_XL	21,789,618,976	0.003071	97.724%
Unsloth UD-Q6_K_XL	26,265,362,976	0.001034	98.752%

On this evaluator V1 beats the compared Unsloth Q4, but that is not the useful headline: V1 has about 1.99x the KLD of Unsloth Q5 despite being roughly 2.76 GB larger, and about 5.92x the KLD of Unsloth Q6. This package is a speed-oriented engineering draft and a reproducible starting point for R9V, not a recommendation over Unsloth's quality quants.

Current product limitations

The published R9V user runtime is not complete yet; current speed figures come from the frozen raw-token proof engine.
No OpenAI-compatible API or chat template is qualified for this profile.
Vision and DFlash artifacts are optional package components, but the speed comparison does not exercise them.
The proof engine is specialized for gfx1201 and the exact tensor layout.
The TG run reports 208 attention-pin fallbacks per 256-token sample; this is disclosed in the benchmark record rather than hidden.

See the R9V benchmark report and qualification report.

License and provenance

The model artifact remains under Apache License 2.0. It is derived immediately from Unsloth's pinned Q8_0 GGUF of Meta's Muse Glimmer 30B model and was requantized with llama.cpp/ggml tooling. Exact revisions and hashes are in sources.lock.json; complete attribution and Meta usage-policy guidance are in THIRD_PARTY_NOTICES.md.

Downloads last month: -

GGUF

Model size

28B params

Architecture

muse-glimmer

Hardware compatibility

We're not able to determine the quantization variants.

View all variants

Model tree for Dyluhn/Muse-Glimmer-30B-R9V-V1

Base model

meta-models/Muse-Glimmer-30B

Quantized

(158)

this model