Qwen3.8 2.4T A95B qwen3.8-2.4t-a95b

qwen · qwen

#4 overall

params
2446.2B
arch
moe
context
1024k
license
other open weights
released
Aug 2026
downloads/30d
25.2k

# architecture

attn shared GQA 64q/4kv router top-10 of 512 expert ×512 expert out ×92 layers ctx 1,048,576 2446.2B pool
layers
92
d_model
8192
heads
64q / 4kv
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: #4 — score 96.0 ● low — a single signal; treat with caution

signalweightinput (0–100)
aa_intelligence 0.70 94.3
bench_composite 0.30 100.0

benchmark panel evidence:

complete panelscore (0–100)
aa-coding-v1100.0

Coding: provisional — score 95.6 ● 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_coding 0.30 95.6
aider_polyglot 0.30 —
swe_bench_verified 0.40 —

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

full methodology

# trend

OGM score · last 75 days
OGM score over 75 days
overall rank (up is better)
Overall rank over 75 days

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

# benchmarks

benchmarkscoresourcedate
AA Coding Index 71.9 Artificial Analysis —
AA Intelligence Index 39.9 Artificial Analysis —

# where to run

providerquantctx$/M in$/M out$/M cacheprice srctpsuptime✓
Aihubmix 1000k $2.00$6.00— via litellm — —
Alibabaunknown 1000k $2.00$6.00$0.25 via openrouter — 100.0%
DeepInfra 256k $2.00$6.00— via hfrouter 74 —
Novitaunknown 1000k $2.00$6.00— novita — 100.0% ✓
SiliconFlowfp8 1024k $2.00$6.00$0.25 openrouter — 100.0% ✓
Togetherunknown 1010k $2.00$6.00$0.25 via openrouter — 99.8%
Veniceunknown 256k $2.50$7.50— venice — 98.9% ✓

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
openrouter
qwen/qwen3.8-2.4t-a95b