Recipe
glm52-mxfp8-nvfp4-nf3-rtxpro6000-vllm-tp4
glm52-mxfp8-nvfp4-nf3-rtxpro6000-vllm-tp4Immutable TP4/DCP4 deployment for the GLM-5.2 MXFP8/NVFP4/NF3 hybrid checkpoint. Recovered Pop evidence pins the image, checkpoint, vLLM, B12X, and FlashInfer and proves coherent long-context generation, but a current exact-config C1/C2/C4 sweep is still required before promotion.
Record
- Status
- candidate
- Source
- 0xsero
- Engine
- vllm
- Engine version
- c382f1d28d5be2f867c216609408bdb424d6049a+b12x-e44cb77777a075790ebe9f7aa9f225d073aea109
- Graph
- full-and-piecewise
- Accelerators
- 4
- Tensor parallel
- 4
- Context tokens
- 180,007
- chat
- yes
- reasoning
- yes
- tools
- yes
- vision
- no
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.
Compose configuration
vllm
Container · candidate · docker compose
Candidate evidence — not a Run contract
- Compose file
compose.yml- Port
- 8000
Environment
| Variable | Value |
|---|---|
CUDA_VISIBLE_DEVICES | 0,1,2,3 |
CUTE_DSL_ARCH | sm_120a |
HYBRID_KEPT | b12x_nf3 |
HYBRID_NF3 | b12x_nf3 |
VLLM_USE_B12X_MOE | 1 |
VLLM_USE_B12X_SPARSE_INDEXER | 1 |
Measured speed
| Concurrency | Context | Prefill | Decode | TTFT ms | Status | Sweep |
|---|---|---|---|---|---|---|
| 1 | 180,007 | — | 58.2 | — | historical | glm52-mxfp8-nvfp4-nf3-rtxpro6000-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
- glm52-mxfp8-nvfp4-nf3-rtxpro6000-vllm-tp4
- model instance id
- madeby561-glm-5-2-mxfp8-nvfp4-nf3-hybrid--mxfp8-nvfp4-nf3-hybrid
- recipe source
- 0xsero
- schema version
- local-ai-registry/v1
- speed sweep ids
- glm52-mxfp8-nvfp4-nf3-rtxpro6000-vllm-tp4-sweep
- status
- candidate
capabilities
- chat
- Yes
- reasoning
- Yes
- tools
- Yes
- vision
- No
engine
- graph mode
- full-and-piecewise
- name
- vllm
- version
- c382f1d28d5be2f867c216609408bdb424d6049a+b12x-e44cb77777a075790ebe9f7aa9f225d073aea109
serving
- max context tokens
- 180,007
- tensor parallel
- 4
Provenance & metadata (3)
facts
- serving.kv cache tokens · provenance · captured at
- 2026-08-27T06:04:13.773Z
- serving.max concurrency · provenance · captured at
- 2026-08-27T06:04:13.773Z
metadata
provenance
- captured at
- 2026-08-27T06:04:13.773Z
sources
| captured at | kind | url |
|---|---|---|
| 2026-08-27T06:04:13.773Z | normalized-recipe | github.com/0xSero/local-ai-registry ↗ |