Recipe

qwen3-6-27b-fp8-rtx-3090-ti-24gb-vllm-tp2

qwen3-6-27b-fp8-rtx-3090-ti-24gb-vllm-tp2

Record

Status
candidate
Source
localmaxxing
Engine
vllm
Engine version
0.19.1 + MTP n=3
Accelerators
2
Tensor parallel
2
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/Qwen/Qwen3.6-27B-FP8
Repository
Qwen/Qwen3.6-27B-FP8
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

vllm

Evidence only · candidate · reference

Candidate evidence — not a Run contract

Source
https://www.localmaxxing.com/en/runs/cmogs42490005lb04f8a8wm4k

No tokenized launch fields. This is measured or documented compatibility, not a Docker launch.

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
1131,07270.3465.4historicalqwen3-6-27b-fp8-rtx-3090-ti-24gb-vllm-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
qwen3-6-27b-fp8-rtx-3090-ti-24gb-vllm-tp2
model instance id
qwen-qwen3-6-27b-fp8--fp8--33492
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
qwen3-6-27b-fp8-rtx-3090-ti-24gb-vllm-tp2-sweep
status
candidate

capabilities

draft launch

accelerator backend
nvidia
arguments
--model, Qwen/Qwen3.6-27B-FP8, --tensor-parallel-size, 2, --host, 0.0.0.0, --port, 8000
container port
8,000
host port
8,000
image
vllm/vllm-openai@sha256:0a51ea5b4ae2dc5d81890e5173f54203d2a3ae0cfffe51b8fd2afd4391bfd967
ipc
host
kind
docker
shm size
16g
synthesized · generated at
2026-08-31T22:12:17Z
synthesized · image provenance
deepseek-fp8-rtx-pro-6000-blackwell-96gb-vllm-tp1
synthesized · template
vllm-openai-v1

mounts

read onlytarget
No/root/.cache/huggingface

engine

name
vllm
version
0.19.1 + MTP n=3

serving

tensor parallel
2
Provenance & metadata (3)

facts

capabilities.chat · provenance · captured at
2026-08-27T06:04:13.773Z
capabilities.chat · reason capability-not-verified
capabilities.chat · state unknown
capabilities.reasoning · provenance · captured at
2026-08-27T06:04:13.773Z
capabilities.reasoning · reason capability-not-verified
capabilities.reasoning · state unknown
capabilities.tools · provenance · captured at
2026-08-27T06:04:13.773Z
capabilities.tools · reason capability-not-verified
capabilities.tools · state unknown
capabilities.vision · provenance · captured at
2026-08-27T06:04:13.773Z
capabilities.vision · reason capability-not-verified
capabilities.vision · state unknown
description · provenance · captured at
2026-08-27T06:04:13.773Z
description · reason not-observed
description · state unknown
engine.graph mode · provenance · captured at
2026-08-27T06:04:13.773Z
engine.graph mode · reason runtime-detail-not-published
engine.graph mode · state unknown
metadata.localmaxxing.engine flags.attentionBackend · provenance · captured at
2026-08-27T06:04:13.773Z
metadata.localmaxxing.engine flags.attentionBackend · reason not-observed
metadata.localmaxxing.engine flags.attentionBackend · state unknown
metadata.localmaxxing.engine flags.flashAttn · provenance · captured at
2026-08-27T06:04:13.773Z
metadata.localmaxxing.engine flags.flashAttn · reason not-observed
metadata.localmaxxing.engine flags.flashAttn · state unknown
metadata.localmaxxing.engine flags.gpuLayers · provenance · captured at
2026-08-27T06:04:13.773Z
metadata.localmaxxing.engine flags.gpuLayers · reason not-observed
metadata.localmaxxing.engine flags.gpuLayers · state unknown
serving.kv cache tokens · provenance · captured at
2026-08-27T06:04:13.773Z
serving.kv cache tokens · reason kv-cache-capacity-not-published
serving.kv cache tokens · state unknown
serving.max concurrency · provenance · captured at
2026-08-27T06:04:13.773Z
serving.max concurrency · reason not-observed
serving.max concurrency · state unknown
serving.max context tokens · provenance · captured at
2026-08-27T06:04:13.773Z
serving.max context tokens · reason not-observed
serving.max context tokens · state unknown

metadata

localmaxxing · base lineage
known
localmaxxing · engine flags · kvCacheDtype
fp8
localmaxxing · engine flags · mtpEnabled
Yes
localmaxxing · engine flags · specDecoding
Yes
localmaxxing · engine flags · tensorParallel
2
localmaxxing · hardware label
2x RTX 3090 Ti 24GB
localmaxxing · notes
Qwen 3.6 27B FP8 dense + MTP n=3 speculative decoding (FP8 weights ship native mtp.layers.0.* — verified 2026-04-27, mtp_num_hidden_layers=1 in text_config). vLLM 0.19.1 --speculative-config '{"method":"mtp","num_speculative_tokens":3}'. Dual RTX 3090 (Ti slot 1 Gen4 x16 + Dell slot 2 Gen4 x4, hardware-detected as same gpuName for dedupe). Bench: vllm bench serve openai-chat random 512/512 num-prompts 20 concurrency 1 --ignore-eos --extra-body chat_template_kwargs.enable_thinking=false. MTP acceptance 55.02% / avg length 2.65 / per-position P0=75%/P1=54%/P2=36%. TPOT 13.34 ms. **1.59× output throughput vs prior FP8 single-stream (44.17 t/s, no spec decode)**, zero quality cost (same FP8 weights, just MTP draft path). Tool calls verified working (qwen3_coder parser + qwen3 reasoning parser). Thinking-off is the production default — bench verified that thinking-on causes 'silent finish' (model exhausts max_tokens drafting in reasoning channel without emitting visible content) on 5/6 prompt categories.
localmaxxing · revision unpinned
Yes
localmaxxing · run id
cmogs42490005lb04f8a8wm4k

provenance

captured at
2026-08-27T06:04:13.773Z

sources

captured atkindurl
2026-08-27T06:04:13.773Znormalized-recipewww.localmaxxing.com/en/runs/cmogs42490005lb04f8a8wm4k