Recipe
nvidia-nemotron-3-5-lightning-30b-a3b-bf16-nvfp4-rtx-3090-24gb-llama-cpp-tp1
nvidia-nemotron-3-5-lightning-30b-a3b-bf16-nvfp4-rtx-3090-24gb-llama-cpp-tp1Observed LocalMaxxing leaderboard run. Evidence for compatibility, not an executable launch contract.
Record
- Status
- candidate
- Source
- localmaxxing
- Engine
- llama.cpp
- Engine version
- 030ebb558 + PR 26623 @ 000d8ecd4 + local classification fix 94577ad1
- Accelerators
- 1
- Tensor parallel
- 1
- Context tokens
- 8,192
- Max concurrency
- 1
- chat
- unknown
- reasoning
- unknown
- tools
- unknown
- vision
- unknown
Hugging Face model card
Identity
https://huggingface.co/ggml-org/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF- Repository
- ggml-org/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF
- 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
llama.cpp
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.
/home/aliii/opt/llama.cpp-mtp-rollback/build/bin/llama-server--model/home/aliii/models/nemotron-nvfp4/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4.gguf--aliasnemotron-nvfp4-complete-rollback-n7--host127.0.0.1--port8012--ctx-size8192--parallel1--metrics--batch-size512--ubatch-size256--cache-type-kq8_0--cache-type-vq8_0--flash-attnon--jinja--n-gpu-layers999--fitoff--spec-typedraft-mtp--spec-draft-n-max7--threads16--threads-batch16
| Flag | Value |
|---|---|
--model | /home/aliii/models/nemotron-nvfp4/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4.gguf |
--alias | nemotron-nvfp4-complete-rollback-n7 |
--host | 127.0.0.1 |
--port | 8012 |
--ctx-size | 8192 |
--parallel | 1 |
--batch-size | 512 |
--ubatch-size | 256 |
--cache-type-k | q8_0 |
--cache-type-v | q8_0 |
--flash-attn | on |
--n-gpu-layers | 999 |
--fit | off |
--spec-type | draft-mtp |
--spec-draft-n-max | 7 |
--threads | 16 |
--threads-batch | 16 |
Measured speed
| Concurrency | Context | Prefill | Decode | TTFT ms | Status | Sweep |
|---|---|---|---|---|---|---|
| 1 | 8,192 | 1,951.8 | 353.7 | 262.3 | observed | nvidia-nemotron-3-5-lightning-30b-a3b-bf16-nvfp4-rtx-3090-24gb-llama-cpp-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-3090-24gb
- id
- nvidia-nemotron-3-5-lightning-30b-a3b-bf16-nvfp4-rtx-3090-24gb-llama-cpp-tp1
- model instance id
- ggml-org-nvidia-nemotron-3-5-lightning-30b-a3b-gguf--nvfp4
- recipe source
- localmaxxing
- schema version
- local-ai-registry/v1
- speed sweep ids
- nvidia-nemotron-3-5-lightning-30b-a3b-bf16-nvfp4-rtx-3090-24gb-llama-cpp-tp1-sweep
- status
- candidate
capabilities
draft launch
- accelerator backend
- nvidia
- arguments
- -hf, ggml-org/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF, --n-gpu-layers, 999, --host, 0.0.0.0, --port, 8080, -c, 8192
- container port
- 8,080
- environment · LLAMA CACHE
- /root/.cache/huggingface
- host port
- 8,080
- image
- ghcr.io/ggml-org/llama.cpp:server-cuda12-b10481@sha256:b2497f8834f5ecb4e38530f6bf2734b8e0be107f0f0857e259672d1cb85b71c2
- ipc
- host
- kind
- docker
- shm size
- 16g
- synthesized · generated at
- 2026-08-31T22:12:17Z
- synthesized · image provenance
- gemma-4-12b-q4-k-m-rtx-3060-12gb-llama-cpp-tp1
- synthesized · template
- llama-cpp-server-v1
mounts
| read only | target |
|---|---|
| No | /root/.cache/huggingface |
engine
- name
- llama.cpp
- version
- 030ebb558 + PR 26623 @ 000d8ecd4 + local classification fix 94577ad1
serving
- max concurrency
- 1
- max context tokens
- 8,192
- 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 3090
- localmaxxing · notes
- Imported from immutable run 20260812T004722Z: 2 warmups + 7 scored serial requests; exact 512-token array; 1024 output; temperature 0; prompt cache off; median server decode 353.7255 tok/s (353.3801 timer-corrected), client output/wall 324.0063 tok/s; 100% NextN MTP draft acceptance. Highly predictable decode workload, not arbitrary-prose throughput. Complete recurrent rollback PR 26623 plus local memory-classification fix. Exact NVFP4 model SHA 7827805ae9f2d20cc71e46bf05d9cb045e222d3fa0429363c324bbf6d3cab959; 12/12 semantic canaries passed separately. This is official ggml-org NVFP4, not LM Studio Community Q4_K_M.
- localmaxxing · run id
- cmspe7wa500pwmp01c2szra70
- 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/cmspe7wa500pwmp01c2szra70 ↗ |