Qwen3 235B A22B qwen3-235b-a22b

qwen · qwen hybrid-reasoning flagship

unranked — no quality signals yet

params
235B (A22B)
arch
moe
context
128k
license
apache-2.0 open weights
released
Apr 2025
reasoning
yes
train compute
4.8e+24 FLOP
downloads/30d
337.7k

# architecture

attn shared · reasoning GQA 64q/4kv router top-8 of 128 expert ×128 expert out ×94 layers ctx 131,072 235B pool · A22B/token
layers
94
d_model
4096
heads
64q / 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 57.1 ● 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 57.1

benchmark panel evidence:

complete panelscore (0–100)
aider-polyglot-v154.3
aime-2025-v160.0

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

Coding: provisional — score 81.8 ● 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_coding 0.30
aider_polyglot 0.30 81.8
swe_bench_verified 0.40

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
Aider Polyglot 59.6 Aider May 2025
Aider 0.6 / 1 LLM Stats
AIME 2024 0.9 / 1 LLM Stats
AIME 2025 0.8 / 1 LLM Stats
Arena-hard 1.0 / 1 LLM Stats
Bbh 0.9 / 1 LLM Stats
Bfcl 0.7 / 1 LLM Stats

# where to run

providerquantctx$/M in$/M out$/M cacheprice srctpsuptime
Featherless via hfrouter
DeepInfra 40k $0.20$0.60$0.20 via requesty
Nebius 256k $0.20$0.60 via litellm
Nscale 32k $0.20$0.60 via hfrouter 16
Novita 40k $0.20$0.80 via hfrouter 11
Alibabaunknown 128k $0.46$1.82 via openrouter 100.0%
Hyperbolic 128k $2.00$2.00 via litellm

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
qwen3-235b-a22b
aider
Qwen3 235B A22B diff, no think, Alibaba API
arena
Qwen3-235B-A22B
openrouter
qwen/qwen3-235b-a22b