Recipe
muse-glimmer-30b-ud-q3-k-xl-rtx-3080-10gb-llama-cpp-tp1
muse-glimmer-30b-ud-q3-k-xl-rtx-3080-10gb-llama-cpp-tp1Observed LocalMaxxing leaderboard run. Evidence for compatibility, not an executable launch contract.
Record
- Status
- candidate
- Source
- localmaxxing
- Engine
- llama.cpp
- Engine version
- b10472
- Accelerators
- 1
- Tensor parallel
- 1
- Context tokens
- 16,384
- Max concurrency
- 1
- chat
- unknown
- reasoning
- unknown
- tools
- unknown
- vision
- unknown
Hugging Face model card
Identity
https://huggingface.co/unsloth/Muse-Glimmer-30B-GGUF- Repository
- unsloth/Muse-Glimmer-30B-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.
llama-server-mMuse-Glimmer-30B-UD-Q3_K_XL.gguf-c16384-ngl40-faon-ctkq8_0-ctvq8_0--parallel1--jinja--spec-typenone
| Flag | Value |
|---|---|
-m | Muse-Glimmer-30B-UD-Q3_K_XL.gguf |
-c | 16384 |
-ngl | 40 |
-fa | on |
-ctk | q8_0 |
-ctv | q8_0 |
--parallel | 1 |
--spec-type | none |
Measured speed
| Concurrency | Context | Prefill | Decode | TTFT ms | Status | Sweep |
|---|---|---|---|---|---|---|
| 1 | 16,384 | 891.3 | 13.8 | 2,810.9 | observed | muse-glimmer-30b-ud-q3-k-xl-rtx-3080-10gb-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-3080-10gb
- id
- muse-glimmer-30b-ud-q3-k-xl-rtx-3080-10gb-llama-cpp-tp1
- model instance id
- unsloth-muse-glimmer-30b-gguf--ud-q3-k-xl
- recipe source
- localmaxxing
- schema version
- local-ai-registry/v1
- speed sweep ids
- muse-glimmer-30b-ud-q3-k-xl-rtx-3080-10gb-llama-cpp-tp1-sweep
- status
- candidate
capabilities
draft launch
- accelerator backend
- nvidia
- arguments
- -hf, unsloth/Muse-Glimmer-30B-GGUF, --n-gpu-layers, 999, --host, 0.0.0.0, --port, 8080, -c, 16384
- 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
- b10472
serving
- max concurrency
- 1
- max context tokens
- 16,384
- 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 · hardware label
- RTX 3080
- localmaxxing · notes
- Muse-Glimmer-30B UD-Q3_K_XL (12.44 GiB), 52 layers. Does NOT fit 10GB: 40 layers on GPU, 12 on CPU (planned offload, not driver paging - shared host mem stays at 124 MiB). Those 12 CPU layers cost dearly: 13.78 tok/s, and PCIe rx peaks at 20.3 GB/s streaming them every token. ngl 36 measured 9.83 tok/s, so each GPU layer is worth ~1 tok/s here; ngl 40 leaves 632 MiB headroom and is the practical ceiling. Medians of 2 reps. If you have 10GB, a smaller fully-resident quant beats this by 5x. Headless 3080 (both displays on the 7900X iGPU; desktop holds ~64 MiB on the card). Workload: single non-repetitive synthetic log-summarisation prompt, 256 output tokens, temp 0, cache_prompt off, medians of 3 reps (2 where noted). Prefill/decode from llama-server timings; TTFT measured client-side on a streaming request. Prompt token counts differ per model because tokenizers differ. llama.cpp b10472 (60eeeb608), Windows 11 / CUDA 12.4, driver 591.86, PCIe 4.0 x16, stock 370 W.
- localmaxxing · observed command
- llama-server -m Muse-Glimmer-30B-UD-Q3_K_XL.gguf -c 16384 -ngl 40 -fa on -ctk q8_0 -ctv q8_0 --parallel 1 --jinja --spec-type none
- localmaxxing · run id
- cmsy7g9830ct8ms014824moau
- localmaxxing · tokenized · arguments
- llama-server, -m, Muse-Glimmer-30B-UD-Q3_K_XL.gguf, -c, 16384, -ngl, 40, -fa, on, -ctk, q8_0, -ctv, q8_0, --parallel, 1, --jinja, --spec-type, none
- 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/cmsy7g9830ct8ms014824moau ↗ |