models · qwen

Qwen3.6-35B-A3B

image-text-to-text model harvested from Hugging Face (trending). Catalogued at Quarantine — unverified, pending attestation.

Compare Verify

What this registry has checked

Quarantine. We hold the publisher’s own published hash for these weights and have recorded it against this entry. We have not downloaded and hashed the weights ourselves, so the hash is the publisher’s claim, not our measurement.

Method	hf-published-sha256

Re-check this yourself → Source → Full record →

What is recorded

Publisher	qwen
Origin	Recorded as a covered nation under 10 U.S.C. § 4872(f) — the PRC, Russia, Iran or North Korea. The registry did not record which.
Licence	apache-2.0
Parameters	35,951,822,704 (36B)
Context window	262,144 tokens
Modality	text+image+video->text
Native precision	BF16
Pulls recorded here	5,129,809
Downloads reported upstream	3,654,078

Reported scores

GPQA Diamond	84.1
IFBench	64.4
Humanity's Last Exam	22.2

Reported by a third party and recorded here, not re-run by us. Source: artificialanalysis.ai/api/v2.

Measured on real hardware

None of these runs are ours. Each figure is the reporter’s, linked to their own evidence, under their own trust label. The number shown is single-stream decode — one user, concurrency 1 — because aggregate throughput across many users is a different quantity and reads as far faster than anyone will see.

Reported to local-ai-registry

Hardware	Engine	Build measured	Decode, 1 stream	Label	Reported by
Intel Arc Pro B70	vllm	GPTQ-Int4	1140 tok/s @ 16,384	candidate	localmaxxing
Radeon RX 7900 XTX	hipfire	MQ4R	495 tok/s @ 4,096	candidate	localmaxxing
GeForce RTX 5090 ×2 devices	vllm	safetensors NVFP4	341 tok/s @ 32,768	candidate	localmaxxing
GeForce RTX 5090	llama.cpp	Q5_K_M	265 tok/s @ 2,048	candidate	localmaxxing
GeForce RTX 5090	llama.cpp	Unsloth-Dynamic-Q4_K_M	262 tok/s @ 32,768	candidate	localmaxxing
GeForce RTX 5090	llama.cpp	safetensors Q4_K_M	231 tok/s @ 4,096	candidate	localmaxxing
GeForce RTX 5090	llama.cpp	UD-Q6_K	220 tok/s @ 4,096	candidate	localmaxxing
GeForce RTX 4090	llama.cpp	Q4_K_M	214 tok/s @ 32,768	candidate	localmaxxing
GeForce RTX 5080	llama.cpp	APEX-MTP-I-Mini	188 tok/s @ 4,096	candidate	localmaxxing
GeForce RTX 5090	ollama	safetensors Q4_K_M	176 tok/s @ 32,768	candidate	localmaxxing

From local-ai-registry (MIT), commit 124959d: 384 runs across 48 machines, 12 validated — model revision and runtime pinned, launch accepted — and 372 candidate, their label for useful evidence without a reproducible-launch promise. 288 more not listed. 26 report figures that contradict each other — prefill below decode, which parallel prefill essentially never is — and are left to the source rather than ranked here.

Measured by local.ai

Hardware	Build measured	Decode @ 8K	Memory @ 8K	Decode mode
NVIDIA GeForce RTX 5090 32GB	iq1m	268 tok/s	13.9 GB	ordinary
NVIDIA GeForce RTX 5090 32GB	iq2xxs	268 tok/s	14.5 GB	ordinary
NVIDIA GeForce RTX 5090 32GB	UD-IQ2_M	259 tok/s	15.0 GB	ordinary
NVIDIA GeForce RTX 5090 32GB	UD-Q2_K_XL	257 tok/s	15.6 GB	ordinary
NVIDIA RTX PRO 6000 Blackwell 96GB	UD-Q2_K_XL	255 tok/s	18.5 GB	ordinary
NVIDIA RTX PRO 6000 Blackwell 96GB	UD-Q4_K_S	251 tok/s	26.9 GB	ordinary
NVIDIA RTX PRO 6000 Blackwell 96GB	q3km	250 tok/s	22.7 GB	ordinary
NVIDIA RTX PRO 6000 Blackwell 96GB	UD-Q3_K_XL	250 tok/s	22.8 GB	ordinary

Measured by local.ai, not by us: 382 runs, 374 more not listed. Each names its engine pinned by image digest and the exact serve command, in the full record.

Elsewhere in the registry

All models → Browse with filters → What will run on my hardware? →

Registry record as of 2026-09-01. Identity, licence and parameter count are properties of a release and do not move; tier can, and the verify link above re-checks it against the live registry rather than this page.