Local AI Registry

Recipe

ling-2-6-flash-modelopt-fp4-experts-only-b16-rtx-pro-6000-blackwell-96gb-sglang-tp1

ling-2-6-flash-modelopt-fp4-experts-only-b16-rtx-pro-6000-blackwell-96gb-sglang-tp1
JSON APIFind compatible recipes

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

Record

Status
candidate
Source
localmaxxing
Engine
sglang
Engine version
unknown
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/inclusionAI/Ling-2.6-flash
Repository
inclusionAI/Ling-2.6-flash
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

sglang

Evidence only · candidate · reference

Candidate evidence — not a Run contract

Source
https://www.localmaxxing.com/en/runs/cmowsfxrj00bso2014kgqb6l8

Observed source tokens

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

  1. /home/espen/.local/share/rtxbench/venvs/sglang/bin/python
  2. -m
  3. sglang.launch_server
  4. --model-path
  5. /LINUX/Models/inclusionAI/Ling-2.6-flash-NVFP4-EXPERTS-ONLY-c64s512-b16
  6. --served-model-name
  7. ling-2-6-flash-nvfp4-experts-b16
  8. --host
  9. 127.0.0.1
  10. --port
  11. 8001
  12. --context-length
  13. 32768
  14. --mem-fraction-static
  15. 0.86
  16. --trust-remote-code
  17. --quantization
  18. modelopt_fp4
  19. --enable-metrics
  20. --chat-template
  21. /home/espen/Code/vllm-studio/data/templates/ling-copilot-stable-chat_template.jinja
  22. --tool-call-parser
  23. qwen
  24. --chunked-prefill-size
  25. 2048
  26. --disable-radix-cache
  27. --attention-backend
  28. triton
  29. --decode-attention-backend
  30. flashinfer
  31. --moe-runner-backend
  32. auto
  33. --fp4-gemm-backend
  34. flashinfer_cutlass
  35. --cuda-graph-max-bs
  36. 1
  37. --cuda-graph-bs
  38. 1
  39. --skip-server-warmup
  40. --max-running-requests
  41. 1
FlagValue
-msglang.launch_server
--model-path/LINUX/Models/inclusionAI/Ling-2.6-flash-NVFP4-EXPERTS-ONLY-c64s512-b16
--served-model-nameling-2-6-flash-nvfp4-experts-b16
--host127.0.0.1
--port8001
--context-length32768
--mem-fraction-static0.86
--quantizationmodelopt_fp4
--chat-template/home/espen/Code/vllm-studio/data/templates/ling-copilot-stable-chat_template.jinja
--tool-call-parserqwen
--chunked-prefill-size2048
--attention-backendtriton
--decode-attention-backendflashinfer
--moe-runner-backendauto
--fp4-gemm-backendflashinfer_cutlass
--cuda-graph-max-bs1
--cuda-graph-bs1
--max-running-requests1

Environment

VariableValue
CUDA_DEVICE_ORDERPCI_BUS_ID
CUDA_VISIBLE_DEVICES1

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
132,7688,159.484.13,953.2observedling-2-6-flash-modelopt-fp4-experts-only-b16-rtx-pro-6000-blackwell-96gb-sglang-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
ling-2-6-flash-modelopt-fp4-experts-only-b16-rtx-pro-6000-blackwell-96gb-sglang-tp1
model instance id
inclusionai-ling-2-6-flash--modelopt-fp4-experts-only-b16
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
ling-2-6-flash-modelopt-fp4-experts-only-b16-rtx-pro-6000-blackwell-96gb-sglang-tp1-sweep
status
candidate

capabilities

draft launch

accelerator backend
nvidia
arguments
python3, -m, sglang.launch_server, --model-path, inclusionAI/Ling-2.6-flash, --tp, 1, --host, 0.0.0.0, --port, 30000, --context-length, 32768
container port
30,000
host port
30,000
image
lmsysorg/sglang:dev-cu13@sha256:6cd4635214f279e0a43019f88e3120d407567640a58aa7dcc0085e3d91402cc4
ipc
host
kind
docker
shm size
16g
synthesized · generated at
2026-08-31T22:12:17Z
synthesized · image provenance
gemma-4-12b-it-nvfp4-rtxpro6000-sglang-tp1
synthesized · template
sglang-launch-server-v1

mounts

read onlytarget
No/root/.cache/huggingface

engine

name
sglang
version
unknown

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
Automated 32k/64k Pro 6000 batch benchmark. Local model: Ling 2.6 Flash NVFP4 Experts Only b16 at /LINUX/Models/inclusionAI/Ling-2.6-flash-NVFP4-EXPERTS-ONLY-c64s512-b16.
localmaxxing · observed command
CUDA_DEVICE_ORDER=PCI_BUS_ID CUDA_VISIBLE_DEVICES=1 /home/espen/.local/share/rtxbench/venvs/sglang/bin/python -m sglang.launch_server --model-path /LINUX/Models/inclusionAI/Ling-2.6-flash-NVFP4-EXPERTS-ONLY-c64s512-b16 --served-model-name ling-2-6-flash-nvfp4-experts-b16 --host 127.0.0.1 --port 8001 --context-length 32768 --mem-fraction-static 0.86 --trust-remote-code --quantization modelopt_fp4 --enable-metrics --chat-template /home/espen/Code/vllm-studio/data/templates/ling-copilot-stable-chat_template.jinja --tool-call-parser qwen --chunked-prefill-size 2048 --disable-radix-cache --attention-backend triton --decode-attention-backend flashinfer --moe-runner-backend auto --fp4-gemm-backend flashinfer_cutlass --cuda-graph-max-bs 1 --cuda-graph-bs 1 --skip-server-warmup --max-running-requests 1
localmaxxing · run id
cmowsfxrj00bso2014kgqb6l8
localmaxxing · tokenized · arguments
/home/espen/.local/share/rtxbench/venvs/sglang/bin/python, -m, sglang.launch_server, --model-path, /LINUX/Models/inclusionAI/Ling-2.6-flash-NVFP4-EXPERTS-ONLY-c64s512-b16, --served-model-name, ling-2-6-flash-nvfp4-experts-b16, --host, 127.0.0.1, --port, 8001, --context-length, 32768, --mem-fraction-static, 0.86, --trust-remote-code, --quantization, modelopt_fp4, --enable-metrics, --chat-template, /home/espen/Code/vllm-studio/data/templates/ling-copilot-stable-chat_template.jinja, --tool-call-parser, qwen, --chunked-prefill-size, 2048, --disable-radix-cache, --attention-backend, triton, --decode-attention-backend, flashinfer, --moe-runner-backend, auto, --fp4-gemm-backend, flashinfer_cutlass, --cuda-graph-max-bs, 1, --cuda-graph-bs, 1, --skip-server-warmup, --max-running-requests, 1
localmaxxing · tokenized · environment · CUDA DEVICE ORDER
PCI_BUS_ID
localmaxxing · tokenized · environment · CUDA VISIBLE DEVICES
1
localmaxxing · tokenized · fidelity
faithful

provenance

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

sources

captured atkindurl
2026-08-30T09:10:02Znormalized-recipewww.localmaxxing.com/en/runs/cmowsfxrj00bso2014kgqb6l8 ↗