models · qwen

Qwen3-30B-A3B

Qwen3 compact MoE. 30B total / 3B active. Low latency (285ms TTFT). Hybrid thinking mode for efficient reasoning on-device.

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.

Weight file SHA-256	505256bf25d85362d7933843ab7f888b0dba243e2bc71728ac1f88ac0d724edc
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	30,532,122,624 (30.5B)
Context window	128K
Modality	text->text
Architecture	qwen3_moe
Layers	48
Hidden size	2048
Attention	GQA (32 heads, 4 KV)
Native precision	BF16
Pulls recorded here	890,032
Downloads reported upstream	1,859,450

Reported scores

MMLU-Pro	71
GPQA Diamond	51.5
AIME 2025	21.7
MATH-500	86.3
LiveCodeBench	32.2
IFBench	31.9
Humanity's Last Exam	4.6

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
GeForce RTX 3060	llama.cpp	Q4_K_M	116 tok/s @ 2,048	candidate	localmaxxing
Apple M3 Ultra 512GB	llama.cpp	Q4_K_M	102 tok/s @ 8,192	candidate	localmaxxing
GeForce RTX 3060	ollama	Q4_K_M	90.0 tok/s @ 2,048	candidate	localmaxxing
Apple M1 Max 64GB	ollama	Q4_K_M	62.6 tok/s @ 2,048	candidate	localmaxxing
GeForce RTX 3080	ollama	Q4_K_M	31.0 tok/s @ 2,048	candidate	localmaxxing

From local-ai-registry (MIT), commit 124959d: 5 runs across 4 machines, 0 validated — model revision and runtime pinned, launch accepted — and 5 candidate, their label for useful evidence without a reproducible-launch promise.

Elsewhere in the registry

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

Registry record as of 2026-06-11. 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.