Qwen3 30B A3B qwen3-30b-a3b

qwen · qwen hybrid-reasoning efficient

unranked — no quality signals yet

params
30.5B (A3.3B)
arch
moe
context
128k
license
apache-2.0 open weights
released
Apr 2025
reasoning
yes
downloads/30d
2.2M

# architecture

attn shared · reasoning GQA 32q/4kv router top-8 of 128 expert ×128 expert out ×48 layers ctx 131,072 30.5B pool · A3.3B/token
layers
48
d_model
2048
heads
32q / 4kv
head dim
128
experts
top-8 of 128
vocab
151936
family
qwen3_moe

signal path of one layer, generated from the registry's structured fields — dimension lines quote the model's real numbers; an MoE trace forks at the router, a dense trace runs straight through. Geometry fields come from the repo's config.json.

# why ranked

Overall: provisional — score 25.6 ● low — a single signal; treat with caution

provisional: too few independent signals for a numbered position — the score is shown, but this model sorts after every ranked model.

signalweightinput (0–100)
aa_intelligence 0.70
bench_composite 0.30 25.6

benchmark panel evidence:

complete panelscore (0–100)
aime-2025-v125.6

missing signals are dropped and the remaining weights renormalized — never imputed.

full methodology

# trend

OGM score · last 55 days
OGM score over 55 days

methodology changed during this history window; score movement across that boundary is not model movement. See methodology v5.

# benchmarks

benchmarkscoresourcedate
AIME 2024 0.8 / 1 LLM Stats
AIME 2025 0.7 / 1 LLM Stats
Arena-hard 0.9 / 1 LLM Stats
Bfcl 0.7 / 1 LLM Stats

# where to run

providerquantctx$/M in$/M out$/M cacheprice srctpsuptime
Nebius 32k $0.10$0.30 via litellm
DeepInfrafp8 40k $0.12$0.50 deepinfra 99.4%
Alibabaunknown 128k $0.13$0.52 via openrouter 100.0%

sorted by blended price ((3·input + output) / 4 per 1M) · ✓ = the provider's own catalog confirms the offer · "via …" prices are what the aggregator routing the offer charges, not the provider's own list price

source aliases
aa
qwen3-30b-a3b
arena
Qwen3-30B-A3B
openrouter
qwen/qwen3-30b-a3b