howtospark
models / qwen3-6-35b-a3b

Qwen3.6 35B-A3B

Qwen/Qwen3.6-35B-A3B
params
36.0B
active params
3.0B
sparsity
MoE
attention
Full attention
architecture
qwen3_5_moe
license
apache-2.0
downloads
6.6M
likes
2.4k
hf updated
Apr 24, 2026

Visualize

total size

each column is a layer — lit blocks are the experts that fire per token

bits / param
16-bit
footprint
67.0 GiB
sparks to fit
block ≈
281M params
active paramsdormant expertsghost = precision traded awayblock fill ∝ bits kept · Qwen3.6 35B-A3B35.951822704B params

Eval scores (compare all)

Reported

92.7AIME 2026
85.2MMLU-Pro
86GPQA Diamond
80.4LiveCodeBench
73.4SWE-bench Verified