Recipe
muse-glimmer-30b-fp8-radeon-ai-pro-r9700-32gb-vllm-tp3
muse-glimmer-30b-fp8-radeon-ai-pro-r9700-32gb-vllm-tp3Observed LocalMaxxing leaderboard run. Evidence for compatibility, not an executable launch contract.
Record
- Status
- candidate
- Source
- localmaxxing
- Engine
- vllm
- Engine version
- 0.27.0rc2+rocm724
- Accelerators
- 3
- Tensor parallel
- 3
- Context tokens
- 32,768
- Max concurrency
- 16
- chat
- unknown
- reasoning
- unknown
- tools
- unknown
- vision
- unknown
Hugging Face model card
Identity
https://huggingface.co/meta-models/Muse-Glimmer-30B- Repository
- meta-models/Muse-Glimmer-30B
- Status
- known
- Link type
- Exact Hub repository
Public Hugging Face repository confirmed by the Hub API.
Candidate: useful compatibility or speed evidence. The registry does not offer Run until promotion requirements are met.
Observed configuration
vllm
Evidence only · candidate · reference
Candidate evidence — not a Run contract
Observed source tokens
Mechanical split of the source command. Unverified against the engine CLI. The registry does not offer Run for this recipe.
vllmserveMuse-Glimmer-30B-tp2--quantizationfp8--disable-custom-all-reduce--compilation-config{"cudagraph_mode":"NONE"}--speculative-configdflashk=15
| Flag | Value |
|---|---|
-tp | 2 |
--quantization | fp8 |
--compilation-config | {"cudagraph_mode":"NONE"} |
--speculative-config | dflash |
Measured speed
| Concurrency | Context | Prefill | Decode | TTFT ms | Status | Sweep |
|---|---|---|---|---|---|---|
| 16 | 32,768 | — | 331.5 | 736.1 | observed | muse-glimmer-30b-fp8-radeon-ai-pro-r9700-32gb-vllm-tp3-sweep |
Remaining fields
Identity, launch, related records, and measured speed are shown above. This is the rest of the normalized record.
- hardware count
- 3
- hardware id
- radeon-ai-pro-r9700-32gb
- id
- muse-glimmer-30b-fp8-radeon-ai-pro-r9700-32gb-vllm-tp3
- model instance id
- meta-models-muse-glimmer-30b--fp8
- recipe source
- localmaxxing
- schema version
- local-ai-registry/v1
- speed sweep ids
- muse-glimmer-30b-fp8-radeon-ai-pro-r9700-32gb-vllm-tp3-sweep
- status
- candidate
capabilities
engine
- name
- vllm
- version
- 0.27.0rc2+rocm724
serving
- max concurrency
- 16
- max context tokens
- 32,768
- tensor parallel
- 3
Provenance & metadata (3)
facts
- capabilities.chat · provenance · captured at
- 2026-08-30T09:10:02Z
capabilities.chat · reason capability-not-verified
capabilities.chat · state unknown
- capabilities.reasoning · provenance · captured at
- 2026-08-30T09:10:02Z
capabilities.reasoning · reason capability-not-verified
capabilities.reasoning · state unknown
- capabilities.tools · provenance · captured at
- 2026-08-30T09:10:02Z
capabilities.tools · reason capability-not-verified
capabilities.tools · state unknown
- capabilities.vision · provenance · captured at
- 2026-08-30T09:10:02Z
capabilities.vision · reason capability-not-verified
capabilities.vision · state unknown
- engine.graph mode · provenance · captured at
- 2026-08-30T09:10:02Z
engine.graph mode · reason runtime-detail-not-published
engine.graph mode · state unknown
- serving.kv cache tokens · provenance · captured at
- 2026-08-30T09:10:02Z
serving.kv cache tokens · reason kv-cache-capacity-not-published
serving.kv cache tokens · state unknown
metadata
- localmaxxing · backend
- rocm
- localmaxxing · hardware label
- Radeon AI Pro R9700
- localmaxxing · notes
- Ran on 2 of the 3 R9700s (tp2, x8-slot pair). zen-7.0.14 kernel (7.1.x amdgpu RCCL regression bypassed); FP8 on-the-fly; DFlash spec decode k=15. Corrected resubmission of a run that misreported gpuCount=2.
- localmaxxing · observed command
- vllm serve Muse-Glimmer-30B -tp 2 --quantization fp8 --disable-custom-all-reduce --compilation-config '{"cudagraph_mode":"NONE"}' --speculative-config dflash k=15
- localmaxxing · run id
- cmsp1ut7n00i7mp015640hvon
- localmaxxing · tokenized · arguments
- vllm, serve, Muse-Glimmer-30B, -tp, 2, --quantization, fp8, --disable-custom-all-reduce, --compilation-config, {"cudagraph_mode":"NONE"}, --speculative-config, dflash, k=15
- localmaxxing · tokenized · fidelity
- faithful
provenance
- captured at
- 2026-08-30T09:10:02Z
sources
| captured at | kind | url |
|---|---|---|
| 2026-08-30T09:10:02Z | normalized-recipe | www.localmaxxing.com/en/runs/cmsp1ut7n00i7mp015640hvon ↗ |