Muse Glimmer 30B
serving recipe.

Floors are the minimum the vLLM recipe publishes across the whole replica, from vLLM's own runs. The control plane compares them with its own sizing from the model config.

• Tensor ParallelFits inside one node's interconnect. single_node_tp

• DGX Spark (GB10) · 1 GPU
• GB300 NVL4 · 4 GPUs
• RTX 5090 · 1 GPU
• RTX 5090 · 2 GPUs
• MI300X · 8 GPUs
• MI325X · 8 GPUs
• MI355X · 8 GPUs
• Arc Pro B60 · 4 GPUs
• Arc Pro B70 · 4 GPUs

Wingsmith runs models through NVIDIA Dynamo, so it serves on NVIDIA GPUs. Other vendors are listed as the recipe states them.

vllm serve meta-models/Muse-Glimmer-30B \
  --generation-config auto \
  --enable-auto-tool-choice \
  --tool-call-parser muse_glimmer \
  --reasoning-parser muse_glimmer

Features

• Tool callingOn by defaultATEM (XML-style) tool calls with automatic tool choice. Calls arrive in `message.tool_calls`; do not enable guided JSON — this model does not emit JSON tool calls. It emits one tool call per message: several calls arrive as consecutive assistant messages, not a single multi-element `tool_calls` array.
• ReasoningOn by defaultChannel-scoped chain-of-thought. Surfaces as `message.reasoning` (non-streaming) and `delta.reasoning` (streaming). Effort is controlled by a `Reasoning strength: <low|medium|high|xhigh>` line in the system prompt.
• Speculative decodingOpt-inDFlash block-diffusion draft head — predicts a whole block in one forward, verified in parallel by the target. `num_speculative_tokens: 15` is not a tuning knob: the head has `block_size: 16` and slot 0 re-presents the last accepted token, so 15 is what remains.

1. Register the weights

   In the console, or through the API. Resolving a slug proposes an artifact and writes nothing.

   curl "http://localhost:8000/v1/artifacts/resolve?source=hf:meta-models/Muse-Glimmer-30B"

2. Read the draft

   The draft carries this recipe as upstream evidence: the tested flags, the precision and memory floor, the parallelism layouts, and the hardware vLLM verified, with links back to the recipe. Nothing from it is applied without you, and the draft warns when the plane's own sizing and the published floor differ by a quarter or more.

3. Create the workload

   Choose the engine, device shape and scale. The plane picks the clusters, and each cluster's operator renders the deployment as a Dynamo graph.

4. Give it an endpoint

   One model name for clients, routed by the gateway to deployments on every cluster serving it.

Run the control plane locally →

• How much GPU memory does Muse Glimmer 30B need?The vLLM recipe publishes a minimum of 31 GB across the whole replica for its NVFP4 variant; BF16 72 GB, FP8 40 GB. That is vLLM's own measurement, not a per-GPU figure.
• Which parallelism layouts does the Muse Glimmer 30B recipe support?Tensor Parallel.
• Can Wingsmith serve Muse Glimmer 30B?Wingsmith serves models through NVIDIA Dynamo on clusters with NVIDIA GPUs. Register meta-models/Muse-Glimmer-30B and the plane attaches this vLLM recipe to the draft as evidence: its tested flags, precision and memory floor, for you to review rather than applied on your behalf.

From the vLLM recipe for meta-models/Muse-Glimmer-30B ↗ (snapshot at 4bcc189), Apache 2.0.