GigaChat 3.5 432B A28B gigachat3.5-432b-a28b

ai-sage · gigachat

unranked — no quality signals yet

params
432B (A28B)
arch
moe
license
mit open weights
released
Jul 2026
downloads/30d
730

# architecture

attn shared MHA ×64 router top-8 of 256 +1 shared expert ×256 expert out ×40 layers 432B pool · A28B/token
layers
40
d_model
7168
heads
64q / 64kv
head dim
112
experts
top-8 of 256 + 1 shared
vocab
128256
family
gigachat3_5

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.

# trend

collecting history — trends appear after a few daily runs.

# where to run

no known API providers — download the weights from Hugging Face and run them yourself.

source aliases
hf
ai-sage/GigaChat3.5-432B-A28B-GGUF, ai-sage/GigaChat3.5-432B-A28B-base