Qwen3.5 397B A17B qwen3.5-397b-a17b

qwen · qwen hybrid-reasoning flagship

unranked — no quality signals yet

params
403B (A17B)
arch
moe
context
256k
license
apache-2.0 open weights
released
Feb 2026
reasoning
yes

# architecture

attn shared · reasoning GQA 32q/2kv router top-10 of 512 expert ×512 expert out ×60 layers ctx 262,144 403B pool · A17B/token
layers
60
d_model
4096
heads
32q / 2kv
head dim
256
experts
top-10 of 512
vocab
248320
family
qwen3_5_moe_text

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 20.4 ● 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 20.4

benchmark panel evidence:

complete panelscore (0–100)
aime-2026-v11.4
browsecomp-v139.4

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

full methodology

# trend

OGM score · last 112 days
OGM score over 112 days
overall rank (up is better)
Overall rank over 112 days

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

# benchmarks

benchmarkscoresourcedate
Aa-lcr 0.7 / 1 LLM Stats —
AIME 2026 0.9 / 1 LLM Stats —
Bfcl-v4 0.7 / 1 LLM Stats —
Browsecomp 0.7 / 1 LLM Stats —
Browsecomp-zh 0.7 / 1 LLM Stats —
C-eval 0.9 / 1 LLM Stats —

# where to run

providerquantctx$/M in$/M out$/M cacheprice srctpsuptime✓
Featherless — ——— via hfrouter — —
Nebius — ——— nebius — — ✓
Aihubmix 256k $0.16$0.99— via litellm — —
Alibabaunknown 256k $0.39$2.34— via openrouter — 99.0%
SiliconFlow 256k $0.39$2.34— via modelsdev — —
Chutesfp8 256k $0.45$3.00— chutes — — ✓
DeepInfrafp8 256k $0.45$3.00— deepinfra — 99.3% ✓
Parasailfp8 256k $0.50$3.60$0.30 via openrouter — 99.8%
AtlasCloudfp8 256k $0.55$3.50— atlascloud — 99.4% ✓
DigitalOceanunknown 128k $0.55$3.50$0.11 via openrouter — 95.9%
Phalaunknown 256k $0.55$3.50— phala — 98.8% ✓
GMICloudfp8 256k $0.60$3.60— via openrouter — 92.6%
Novitaunknown 256k $0.60$3.60— novita — 97.8% ✓
StreamLakeunknown 250k $0.60$3.60$0.12 via openrouter — 98.0%
Together 256k $0.60$3.60$0.35 via modelsdev — —
Scaleway — $0.68$4.10— via hfrouter 61 —
OVHcloud 256k $0.71$4.25— ovhcloud 46 — ✓
Veniceunknown 125k $0.75$4.50— venice — 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
qwen3.5-397b-a17b
epoch
Qwen3.5 397B-A17B
openrouter
qwen/qwen3.5-397b-a17b