Recipe
glm-5-2-nvfp4-rtx-pro-6000-blackwell-96gb-vllm-tp4
glm-5-2-nvfp4-rtx-pro-6000-blackwell-96gb-vllm-tp4Observed LocalMaxxing leaderboard run. Evidence for compatibility, not an executable launch contract.
Record
- Status
- candidate
- Source
- localmaxxing
- Engine
- vllm
- Accelerators
- 4
- Tensor parallel
- 4
- Context tokens
- 480,000
- Max concurrency
- 1
- chat
- unknown
- reasoning
- unknown
- tools
- unknown
- vision
- unknown
Hugging Face model card
Identity
https://huggingface.co/madeby561/GLM-5.2-MXFP8-NVFP4-NF3-Hybrid- Repository
- madeby561/GLM-5.2-MXFP8-NVFP4-NF3-Hybrid
- 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.
vllmserve/model--tensor-parallel-size4--decode-context-parallel-size4--kv-cache-dtypenvfp4_ds_mla--attention-backendB12X_MLA_SPARSE--moe-backendb12x--gpu-memory-utilization0.986--max-model-len480000--enable-prefix-caching--enable-chunked-prefill--async-scheduling--speculative-config{method:mtp,num_speculative_tokens:3}
| Flag | Value |
|---|---|
--tensor-parallel-size | 4 |
--decode-context-parallel-size | 4 |
--kv-cache-dtype | nvfp4_ds_mla |
--attention-backend | B12X_MLA_SPARSE |
--moe-backend | b12x |
--gpu-memory-utilization | 0.986 |
--max-model-len | 480000 |
--speculative-config | {method:mtp,num_speculative_tokens:3} |
Measured speed
| Concurrency | Context | Prefill | Decode | TTFT ms | Status | Sweep |
|---|---|---|---|---|---|---|
| 1 | 480,000 | — | 79.6 | — | observed | glm-5-2-nvfp4-rtx-pro-6000-blackwell-96gb-vllm-tp4-sweep |
Remaining fields
Identity, launch, related records, and measured speed are shown above. This is the rest of the normalized record.
- hardware count
- 4
- hardware id
- rtx-pro-6000-blackwell-96gb
- id
- glm-5-2-nvfp4-rtx-pro-6000-blackwell-96gb-vllm-tp4
- model instance id
- madeby561-glm-5-2-mxfp8-nvfp4-nf3-hybrid--nvfp4
- recipe source
- localmaxxing
- schema version
- local-ai-registry/v1
- speed sweep ids
- glm-5-2-nvfp4-rtx-pro-6000-blackwell-96gb-vllm-tp4-sweep
- status
- candidate
capabilities
draft launch
- accelerator backend
- nvidia
- arguments
- --model, madeby561/GLM-5.2-MXFP8-NVFP4-NF3-Hybrid, --tensor-parallel-size, 4, --host, 0.0.0.0, --port, 8000, --max-model-len, 480000
- 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 only | target |
|---|---|
| No | /root/.cache/huggingface |
engine
- name
- vllm
serving
- max concurrency
- 1
- max context tokens
- 480,000
- tensor parallel
- 4
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 · hardware label
- RTX PRO 6000 Blackwell
- localmaxxing · observed command
- vllm serve /model --tensor-parallel-size 4 --decode-context-parallel-size 4 --kv-cache-dtype nvfp4_ds_mla --attention-backend B12X_MLA_SPARSE --moe-backend b12x --gpu-memory-utilization 0.986 --max-model-len 480000 --enable-prefix-caching --enable-chunked-prefill --async-scheduling --speculative-config {method:mtp,num_speculative_tokens:3}
- localmaxxing · run id
- cms2jeldj00n7s201llzdtqbr
- localmaxxing · tokenized · arguments
- vllm, serve, /model, --tensor-parallel-size, 4, --decode-context-parallel-size, 4, --kv-cache-dtype, nvfp4_ds_mla, --attention-backend, B12X_MLA_SPARSE, --moe-backend, b12x, --gpu-memory-utilization, 0.986, --max-model-len, 480000, --enable-prefix-caching, --enable-chunked-prefill, --async-scheduling, --speculative-config, {method:mtp,num_speculative_tokens:3}
- 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/cms2jeldj00n7s201llzdtqbr ↗ |