Recipe
qwen3-6-35b-a3b-ud-q4-k-xl-rtx-3060-12gb-llama-cpp-tp2
qwen3-6-35b-a3b-ud-q4-k-xl-rtx-3060-12gb-llama-cpp-tp2Observed LocalMaxxing leaderboard run. Evidence for compatibility, not an executable launch contract.
Record
- Status
- candidate
- Source
- localmaxxing
- Engine
- llama.cpp
- Engine version
- b10682
- Accelerators
- 2
- Tensor parallel
- 2
- Context tokens
- 90,112
- Max concurrency
- 1
- chat
- unknown
- reasoning
- unknown
- tools
- unknown
- vision
- unknown
Hugging Face model card
Identity
https://huggingface.co/unsloth/Qwen3.6-35B-A3B-GGUF- Repository
- unsloth/Qwen3.6-35B-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.
llama-server-mQwen3.6-35B-A3B-UD-Q4_K_XL.gguf-ngl99--split-modetensor--tensor-split1,1-c90112-ctkq4_0-ctvq4_0--cache-type-k-draftq4_0--cache-type-v-draftq4_0-np1-faon--spec-typedraft-mtp--spec-draft-n-max2
| Flag | Value |
|---|---|
-m | Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf |
-ngl | 99 |
--split-mode | tensor |
--tensor-split | 1,1 |
-c | 90112 |
-ctk | q4_0 |
-ctv | q4_0 |
--cache-type-k-draft | q4_0 |
--cache-type-v-draft | q4_0 |
-np | 1 |
-fa | on |
--spec-type | draft-mtp |
--spec-draft-n-max | 2 |
Environment
| Variable | Value |
|---|---|
GGML_CUDA_ALLREDUCE | internal |
Measured speed
| Concurrency | Context | Prefill | Decode | TTFT ms | Status | Sweep |
|---|---|---|---|---|---|---|
| 1 | 90,112 | — | 129.4 | — | observed | qwen3-6-35b-a3b-ud-q4-k-xl-rtx-3060-12gb-llama-cpp-tp2-sweep |
Remaining fields
Identity, launch, related records, and measured speed are shown above. This is the rest of the normalized record.
- hardware count
- 2
- hardware id
- rtx-3060-12gb
- id
- qwen3-6-35b-a3b-ud-q4-k-xl-rtx-3060-12gb-llama-cpp-tp2
- model instance id
- unsloth-qwen3-6-35b-a3b-gguf--ud-q4-k-xl
- recipe source
- localmaxxing
- schema version
- local-ai-registry/v1
- speed sweep ids
- qwen3-6-35b-a3b-ud-q4-k-xl-rtx-3060-12gb-llama-cpp-tp2-sweep
- status
- candidate
capabilities
engine
- name
- llama.cpp
- version
- b10682
serving
- max concurrency
- 1
- max context tokens
- 90,112
- tensor parallel
- 2
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
- engine.version · provenance · captured at
- 2026-09-01T01:34:02Z
engine.version · reason explicit-engine-version-in-live-run-evidence
engine.version · state known
- 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
- serving.max concurrency · provenance · captured at
- 2026-08-31T23:03:15Z
serving.max concurrency · reason server-capacity-derived-from-source-evidence
serving.max concurrency · state known
metadata
- localmaxxing · batch size
- 1
- localmaxxing · hardware label
- RTX 3060
- localmaxxing · notes
- Dual RTX 3060 12GB, no NVLink. Chat decode tok/s with native MTP (spec-type draft-mtp, n-max 2), thinking off, 256-token code prompt. 88k ctx (90112). Main and draft KV q4_0. Not llama-bench without speculation. llama.cpp b10682.
- localmaxxing · observed command
- GGML_CUDA_ALLREDUCE=internal llama-server -m Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf -ngl 99 --split-mode tensor --tensor-split 1,1 -c 90112 -ctk q4_0 -ctv q4_0 --cache-type-k-draft q4_0 --cache-type-v-draft q4_0 -np 1 -fa on --spec-type draft-mtp --spec-draft-n-max 2
- localmaxxing · run id
- cmteqha6p0034n0016to5zgpe
- localmaxxing · tokenized · arguments
- llama-server, -m, Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf, -ngl, 99, --split-mode, tensor, --tensor-split, 1,1, -c, 90112, -ctk, q4_0, -ctv, q4_0, --cache-type-k-draft, q4_0, --cache-type-v-draft, q4_0, -np, 1, -fa, on, --spec-type, draft-mtp, --spec-draft-n-max, 2
- localmaxxing · tokenized · environment · GGML CUDA ALLREDUCE
- internal
- 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/cmteqha6p0034n0016to5zgpe ↗ |