Local AI Registry

Recipe

hy3-nvfp4-dgx-spark-gb10-128gb-vllm-tp1

hy3-nvfp4-dgx-spark-gb10-128gb-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
0.25.1.dev24+g96bb89286.d20260710
Accelerators
1
Tensor parallel
1
Context tokens
4,096
Max concurrency
1
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/Jiunsong/SuperHY3-abliterated-NVFP4
Repository
Jiunsong/SuperHY3-abliterated-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/cmrjee41802humj01va09q9a1

Observed source tokens

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

  1. vllm
  2. serve
  3. Jiunsong/SuperHY3-abliterated-NVFP4
  4. --served-model-name
  5. superhy3
  6. --host
  7. 0.0.0.0
  8. --port
  9. 8600
  10. --tensor-parallel-size
  11. 2
  12. --distributed-executor-backend
  13. ray
  14. --max-model-len
  15. 4096
  16. --max-num-seqs
  17. 1
  18. --max-num-batched-tokens
  19. 4096
  20. --gpu-memory-utilization
  21. 0.85
  22. --kv-cache-dtype
  23. bfloat16
  24. --speculative-config
  25. {"method":"mtp","num_speculative_tokens":1}
  26. --generation-config
  27. vllm
  28. --trust-remote-code
  29. --enforce-eager
FlagValue
--served-model-namesuperhy3
--host0.0.0.0
--port8600
--tensor-parallel-size2
--distributed-executor-backendray
--max-model-len4096
--max-num-seqs1
--max-num-batched-tokens4096
--gpu-memory-utilization0.85
--kv-cache-dtypebfloat16
--speculative-config{"method":"mtp","num_speculative_tokens":1}
--generation-configvllm

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
14,09667519195.6observedhy3-nvfp4-dgx-spark-gb10-128gb-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
dgx-spark-gb10-128gb
id
hy3-nvfp4-dgx-spark-gb10-128gb-vllm-tp1
model instance id
jiunsong-superhy3-abliterated-nvfp4--nvfp4
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
hy3-nvfp4-dgx-spark-gb10-128gb-vllm-tp1-sweep
status
candidate

capabilities

draft launch

accelerator backend
nvidia
arguments
--model, Jiunsong/SuperHY3-abliterated-NVFP4, --tensor-parallel-size, 1, --host, 0.0.0.0, --port, 8000, --max-model-len, 4096
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
0.25.1.dev24+g96bb89286.d20260710

serving

max concurrency
1
max context tokens
4,096
tensor parallel
1
Provenance & metadata (3)

facts

capabilities.chat · provenance · captured at
2026-08-30T09:26:09Z
capabilities.chat · reason capability-not-verified
capabilities.chat · state unknown
capabilities.reasoning · provenance · captured at
2026-08-30T09:26:09Z
capabilities.reasoning · reason capability-not-verified
capabilities.reasoning · state unknown
capabilities.tools · provenance · captured at
2026-08-30T09:26:09Z
capabilities.tools · reason capability-not-verified
capabilities.tools · state unknown
capabilities.vision · provenance · captured at
2026-08-30T09:26:09Z
capabilities.vision · reason capability-not-verified
capabilities.vision · state unknown
engine.graph mode · provenance · captured at
2026-08-30T09:26:09Z
engine.graph mode · reason runtime-detail-not-published
engine.graph mode · state unknown
serving.kv cache tokens · provenance · captured at
2026-08-30T09:26:09Z
serving.kv cache tokens · reason kv-cache-capacity-not-published
serving.kv cache tokens · state unknown

metadata

localmaxxing · backend
cuda
localmaxxing · hardware label
GB10 Grace Blackwell
localmaxxing · notes
Official LocalMaxxing CLI v0.1.19 remote-endpoint run against commit bc41c8ce4ce0be46b60cc4d728e1b85eee1107a1: 2 untimed warmups + 7 timed streamed iterations, temperature=0, standard CLI prompt, 132 prompt / 256 output tokens, batch=1. Reported values are medians: 19.0 decode tok/s (18.8-19.4, mean 19.06, stddev 0.21), 195.56 ms TTFT, 675 tok/s estimated prefill, and 28.5 total tok/s. Runtime: 2x NVIDIA DGX Spark (GB10), 256 GB aggregate unified memory, TP=2 over direct 200GbE/RoCE (NCCL NET/IB); vLLM 0.25.1.dev24+g96bb89286.d20260710; compressed-tensors MARLIN NVFP4 W4A16 routed experts with BF16 non-experts and BF16 KV; max_model_len=4096, max_num_seqs=1, eager, gpu_memory_utilization=0.85, native MTP spec-1. Session MTP counters after integrity plus benchmark: 896/1555 accepted draft tokens (57.62%). Peak VRAM is the max per node: 101.72 GiB target allocation (~203.44 GiB aggregate); vLLM loaded 84.49-84.50 GiB weights per rank. Pre-submit integrity suite passed 5/5: arithmetic, strict JSON, Korean, Python AST, and exact-format generation; no empty output, invalid UTF-8, or special-token leakage. Model files were full-checksum verified across both nodes. LocalMaxxing hardware memory is reported per node (128 GB); the two-node cluster total is 256 GB.
localmaxxing · observed command
vllm serve Jiunsong/SuperHY3-abliterated-NVFP4 --served-model-name superhy3 --host 0.0.0.0 --port 8600 --tensor-parallel-size 2 --distributed-executor-backend ray --max-model-len 4096 --max-num-seqs 1 --max-num-batched-tokens 4096 --gpu-memory-utilization 0.85 --kv-cache-dtype bfloat16 --speculative-config '{"method":"mtp","num_speculative_tokens":1}' --generation-config vllm --trust-remote-code --enforce-eager
localmaxxing · run id
cmrjee41802humj01va09q9a1
localmaxxing · tokenized · arguments
vllm, serve, Jiunsong/SuperHY3-abliterated-NVFP4, --served-model-name, superhy3, --host, 0.0.0.0, --port, 8600, --tensor-parallel-size, 2, --distributed-executor-backend, ray, --max-model-len, 4096, --max-num-seqs, 1, --max-num-batched-tokens, 4096, --gpu-memory-utilization, 0.85, --kv-cache-dtype, bfloat16, --speculative-config, {"method":"mtp","num_speculative_tokens":1}, --generation-config, vllm, --trust-remote-code, --enforce-eager
localmaxxing · tokenized · fidelity
faithful

provenance

captured at
2026-08-30T09:26:09Z

sources

captured atkindurl
2026-08-30T09:26:09Znormalized-recipewww.localmaxxing.com/en/runs/cmrjee41802humj01va09q9a1 ↗