Local AI Registry

Recipe

qwen3-6-35b-a3b-ud-q6-k-xl-rtx-3090-ti-24gb-llama-cpp-tp2

qwen3-6-35b-a3b-ud-q6-k-xl-rtx-3090-ti-24gb-llama-cpp-tp2
JSON APIFind compatible recipes

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

Record

Status
candidate
Source
localmaxxing
Engine
llama.cpp
Engine version
b1-6217b49 (CUDA)
Accelerators
2
Tensor parallel
2
Context tokens
65,536
Max concurrency
1
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/Qwen/Qwen3.6-35B-A3B
Repository
Qwen/Qwen3.6-35B-A3B
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

Source
https://www.localmaxxing.com/en/runs/cmogsauq70006k40418jb5vob

Observed source tokens

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

  1. llama-server
  2. -m
  3. Qwen3.6-35B-A3B-UD-Q6_K_XL.gguf
  4. --alias
  5. qwen3.6-35b-a3b-q6
  6. -ngl
  7. 999
  8. -fa
  9. 1
  10. --no-mmap
  11. -c
  12. 65536
  13. --cache-reuse
  14. 256
  15. -ctk
  16. q8_0
  17. -ctv
  18. q8_0
  19. --no-context-shift
  20. --parallel
  21. 4
  22. -b
  23. 2048
  24. -ub
  25. 2048
  26. --metrics
  27. --jinja
  28. --chat-template-kwargs
  29. {"enable_thinking":false}
  30. --temp
  31. 1.0
  32. --top-p
  33. 0.95
  34. --top-k
  35. 20
  36. --min-p
  37. 0.0
  38. --host
  39. 0.0.0.0
  40. --port
  41. 8000
FlagValue
-mQwen3.6-35B-A3B-UD-Q6_K_XL.gguf
--aliasqwen3.6-35b-a3b-q6
-ngl999
-fa1
-c65536
--cache-reuse256
-ctkq8_0
-ctvq8_0
--parallel4
-b2048
-ub2048
--chat-template-kwargs{"enable_thinking":false}
--temp1.0
--top-p0.95
--top-k20
--min-p0.0
--host0.0.0.0
--port8000

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
165,536—131.6212.6observedqwen3-6-35b-a3b-ud-q6-k-xl-rtx-3090-ti-24gb-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-3090-ti-24gb
id
qwen3-6-35b-a3b-ud-q6-k-xl-rtx-3090-ti-24gb-llama-cpp-tp2
model instance id
qwen-qwen3-6-35b-a3b--ud-q6-k-xl
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
qwen3-6-35b-a3b-ud-q6-k-xl-rtx-3090-ti-24gb-llama-cpp-tp2-sweep
status
candidate

capabilities

engine

name
llama.cpp
version
b1-6217b49 (CUDA)

serving

max concurrency
1
max context tokens
65,536
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
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 Ti
localmaxxing · notes
Qwen 3.6 35B-A3B MoE (3B active) UD-Q6_K_XL via llama-server (Unsloth Dynamic 2.0). Dual RTX 3090 (Ti slot 1 Gen4 x16 + Dell slot 2 Gen4 x4), default layer split. Bench: vllm bench serve openai-chat random 512/512 num-prompts 20 concurrency 1 --ignore-eos. TPOT 7.20 ms — MoE 3B-active sparse activation gives sharply lower per-token latency than dense FP8 27B (13.34 ms TPOT). Total throughput 263 t/s. Quant choice matches Bosgame M5 production sweet spot (validated against Hermes agent harness — 8/8 tool-call passes, clean format compliance). Thinking off via server-side --chat-template-kwargs '{"enable_thinking":false}'. Use case: HTML/forms/quick-codegen mode in opencode toggle (vllm-mode dense|moe). VRAM 17 GB per card across both = ~34 GB total (Q6_K_XL 32 GB on disk). Prefix caching off, KV q8_0.
localmaxxing · observed command
llama-server -m Qwen3.6-35B-A3B-UD-Q6_K_XL.gguf --alias qwen3.6-35b-a3b-q6 -ngl 999 -fa 1 --no-mmap -c 65536 --cache-reuse 256 -ctk q8_0 -ctv q8_0 --no-context-shift --parallel 4 -b 2048 -ub 2048 --metrics --jinja --chat-template-kwargs '{"enable_thinking":false}' --temp 1.0 --top-p 0.95 --top-k 20 --min-p 0.0 --host 0.0.0.0 --port 8000
localmaxxing · run id
cmogsauq70006k40418jb5vob
localmaxxing · tokenized · arguments
llama-server, -m, Qwen3.6-35B-A3B-UD-Q6_K_XL.gguf, --alias, qwen3.6-35b-a3b-q6, -ngl, 999, -fa, 1, --no-mmap, -c, 65536, --cache-reuse, 256, -ctk, q8_0, -ctv, q8_0, --no-context-shift, --parallel, 4, -b, 2048, -ub, 2048, --metrics, --jinja, --chat-template-kwargs, {"enable_thinking":false}, --temp, 1.0, --top-p, 0.95, --top-k, 20, --min-p, 0.0, --host, 0.0.0.0, --port, 8000
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/cmogsauq70006k40418jb5vob ↗