Qwen3 235B A22B Instruct 2507 qwen3-235b-a22b-instruct-2507

qwen · qwen flagship

unranked — no quality signals yet

params
235.1B (A22B)
arch
moe
context
256k
license
apache-2.0 open weights
released
Jul 2025
train compute
4.8e+24 FLOP

# architecture

attn shared GQA 64q/4kv router top-8 of 128 expert ×128 expert out ×94 layers ctx 262,144 235.1B 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 — ● 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 0.0

benchmark panel evidence:

complete panelscore (0–100)
aime-2025-v10.0

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

full methodology

# trend

OGM score · last 108 days
OGM score over 108 days
overall rank (up is better)
Overall rank over 108 days

methodology changed during this history window; score movement across that boundary is not model movement. See methodology v6.

# benchmarks

benchmarkscoresourcedate
Aider-polyglot 0.6 / 1 LLM Stats —
AIME 2025 0.7 / 1 LLM Stats —
Arc-agi 0.4 / 1 LLM Stats —
Arena-hard-v2 0.8 / 1 LLM Stats —
Bfcl-v3 0.7 / 1 LLM Stats —

# where to run

providerquantctx$/M in$/M out$/M cacheprice srctpsuptime✓
Nebius — ——— nebius — — ✓
GMICloudfp8 256k $0.09$0.35$0.02 via openrouter — —
DeepInfrafp8 256k $0.09$0.55— deepinfra — 96.1% ✓
Novita 128k $0.09$0.58— via litellm — —
Venicefp8 125k $0.15$0.75— venice — 99.0% ✓
Nscale 32k $0.20$0.60— via hfrouter 15 —
Parasailfp8 128k $0.14$0.80$0.05 via openrouter — 100.0%
Amazon Bedrock — $0.22$0.88— bedrock — — ✓
Replicate — $0.26$1.06— via litellm — —
Scaleway — $0.85$2.56— via hfrouter 45 —
Crusoe 256k $3.00$3.00— 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-instruct-2507
bedrock
Qwen3 235B A22B 2507
epoch
Qwen3-235B-A22B (Jul 2025)
llmstats
qwen3-235b-a22b-instruct-2507
openrouter
qwen/qwen3-235b-a22b-2507