Local AI Registry

Recipe

muse-glimmer-30b-nvfp4-rtx-pro-6000-blackwell-96gb-vllm-tp1

muse-glimmer-30b-nvfp4-rtx-pro-6000-blackwell-96gb-vllm-tp1
JSON APIFind compatible recipes

Observed LocalMaxxing leaderboard run. Evidence for compatibility, not an executable launch contract.

Record

Status
candidate
Source
localmaxxing
Engine
vllm
Engine version
0.1.dev1+g075d645af
Accelerators
1
Tensor parallel
1
Context tokens
32,768
Max concurrency
1
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/Preyazz/Muse-Glimmer-30B-NVFP4
Repository
Preyazz/Muse-Glimmer-30B-NVFP4
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

Source
https://www.localmaxxing.com/en/runs/cmsnghgre00e0o001qeoj0m0o

Observed source tokens

Mechanical split of the source command. Unverified against the engine CLI. The registry does not offer Run for this recipe.

    1. cd
    2. {PROJECT_ROOT}
    1. source
    2. .venv-vllm/bin/activate
    1. vllm
    2. serve
    3. {PROJECT_ROOT}/models/preyazz-muse-glimmer-30b-nvfp4
    4. --served-model-name
    5. muse-glimmer
    6. --host
    7. 127.0.0.1
    8. --port
    9. 8000
    10. --tensor-parallel-size
    11. 1
    12. --gpu-memory-utilization
    13. 0.90
    14. --max-model-len
    15. 32768
    16. --max-num-seqs
    17. 8
    18. --max-num-batched-tokens
    19. 32768
    20. --enable-auto-tool-choice
    21. --tool-call-parser
    22. muse_glimmer
    23. --reasoning-parser
    24. muse_glimmer
    25. --generation-config
    26. auto
    27. --speculative-config
    28. {"method":"dflash","model":"{PROJECT_ROOT}/models/meta-muse-glimmer-30b-assistant","num_speculative_tokens":10}
FlagValue
--served-model-namemuse-glimmer
--host127.0.0.1
--port8000
--tensor-parallel-size1
--gpu-memory-utilization0.90
--max-model-len32768
--max-num-seqs8
--max-num-batched-tokens32768
--tool-call-parsermuse_glimmer
--reasoning-parsermuse_glimmer
--generation-configauto
--speculative-config{"method":"dflash","model":"{PROJECT_ROOT}/models/meta-muse-glimmer-30b-assistant","num_speculative_tokens":10}

Environment

VariableValue
HF_HOME{PROJECT_ROOT}/.cache/huggingface
MAX_JOBS4
NVCC_THREADS2
TORCH_CUDA_ARCH_LIST12.0a
VLLM_USE_DEEP_GEMM0
XDG_CACHE_HOME{PROJECT_ROOT}/.cache

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
132,7684,396.9152.222.4observedmuse-glimmer-30b-nvfp4-rtx-pro-6000-blackwell-96gb-vllm-tp1-sweep

Remaining fields

Identity, launch, related records, and measured speed are shown above. This is the rest of the normalized record.

hardware count
1
hardware id
rtx-pro-6000-blackwell-96gb
id
muse-glimmer-30b-nvfp4-rtx-pro-6000-blackwell-96gb-vllm-tp1
model instance id
preyazz-muse-glimmer-30b-nvfp4--nvfp4
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
muse-glimmer-30b-nvfp4-rtx-pro-6000-blackwell-96gb-vllm-tp1-sweep
status
candidate

capabilities

draft launch

accelerator backend
nvidia
arguments
--model, Preyazz/Muse-Glimmer-30B-NVFP4, --tensor-parallel-size, 1, --host, 0.0.0.0, --port, 8000, --max-model-len, 32768
container port
8,000
host port
8,000
image
vllm/vllm-openai@sha256:0a51ea5b4ae2dc5d81890e5173f54203d2a3ae0cfffe51b8fd2afd4391bfd967
ipc
host
kind
docker
shm size
16g
synthesized · generated at
2026-08-31T22:12:17Z
synthesized · image provenance
deepseek-fp8-rtx-pro-6000-blackwell-96gb-vllm-tp1
synthesized · template
vllm-openai-v1

mounts

read onlytarget
No/root/.cache/huggingface

engine

name
vllm
version
0.1.dev1+g075d645af

serving

max concurrency
1
max context tokens
32,768
tensor parallel
1
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
cuda
localmaxxing · hardware label
RTX PRO 6000 Blackwell
localmaxxing · notes
NVFP4 weight-only Marlin target with DFlash 10 tokens, the best confirmed single-stream setting. Vision remained enabled; the same target/drafter vision path passed a real PNG smoke test during this sweep. BS8 683.7 tok/s, alt config for Best BS8: 7 draft tokens, 816.5 tok/s aggregate (102.1 per stream); single 148.1 tok/s
localmaxxing · observed command
cd {PROJECT_ROOT} && source .venv-vllm/bin/activate && MAX_JOBS=4 NVCC_THREADS=2 TORCH_CUDA_ARCH_LIST=12.0a VLLM_USE_DEEP_GEMM=0 HF_HOME={PROJECT_ROOT}/.cache/huggingface XDG_CACHE_HOME={PROJECT_ROOT}/.cache vllm serve {PROJECT_ROOT}/models/preyazz-muse-glimmer-30b-nvfp4 --served-model-name muse-glimmer --host 127.0.0.1 --port 8000 --tensor-parallel-size 1 --gpu-memory-utilization 0.90 --max-model-len 32768 --max-num-seqs 8 --max-num-batched-tokens 32768 --enable-auto-tool-choice --tool-call-parser muse_glimmer --reasoning-parser muse_glimmer --generation-config auto --speculative-config '{"method":"dflash","model":"{PROJECT_ROOT}/models/meta-muse-glimmer-30b-assistant","num_speculative_tokens":10}'
localmaxxing · run id
cmsnghgre00e0o001qeoj0m0o
localmaxxing · tokenized · arguments
vllm, serve, {PROJECT_ROOT}/models/preyazz-muse-glimmer-30b-nvfp4, --served-model-name, muse-glimmer, --host, 127.0.0.1, --port, 8000, --tensor-parallel-size, 1, --gpu-memory-utilization, 0.90, --max-model-len, 32768, --max-num-seqs, 8, --max-num-batched-tokens, 32768, --enable-auto-tool-choice, --tool-call-parser, muse_glimmer, --reasoning-parser, muse_glimmer, --generation-config, auto, --speculative-config, {"method":"dflash","model":"{PROJECT_ROOT}/models/meta-muse-glimmer-30b-assistant","num_speculative_tokens":10}
localmaxxing · tokenized · environment · HF HOME
{PROJECT_ROOT}/.cache/huggingface
localmaxxing · tokenized · environment · MAX JOBS
4
localmaxxing · tokenized · environment · NVCC THREADS
2
localmaxxing · tokenized · environment · TORCH CUDA ARCH LIST
12.0a
localmaxxing · tokenized · environment · VLLM USE DEEP GEMM
0
localmaxxing · tokenized · environment · XDG CACHE HOME
{PROJECT_ROOT}/.cache
localmaxxing · tokenized · fidelity
faithful

localmaxxing · tokenized · steps

01234567
cd{PROJECT_ROOT}UnknownUnknownUnknownUnknownUnknownUnknown
source.venv-vllm/bin/activateUnknownUnknownUnknownUnknownUnknownUnknown
vllmserve{PROJECT_ROOT}/models/preyazz-muse-glimmer-30b-nvfp4--served-model-namemuse-glimmer--host127.0.0.1--port

provenance

captured at
2026-08-30T09:10:02Z

sources

captured atkindurl
2026-08-30T09:10:02Znormalized-recipewww.localmaxxing.com/en/runs/cmsnghgre00e0o001qeoj0m0o ↗