Search · KernelIndex

Skip to content

Search Models Records GPUs Docs

More

Records GPUs Docs

Kernels Serving

⏎

Feed Serving Projects Challenges Contribute Sources Methodology Legal GitHub Atom llms.txta46b010

Kernels Serving

⏎

Trygemm B200 bf16 moe B200 bf16or filter withgpu:B200dtype:bf16shape:[2048,4096]model:deepseek-v3Query syntax

Browse the index

26 rows with ranked runs · 25,513 kernel runssorted byMost indexedRecently active A–Z

All families gemm moe rmsnorm gqa-paged-attention other mla-attention rope gemv

Operation

Family

Published runs

Last run

GQA paged decode h32 kv4 d128 ps1gqa-paged-attention6012026-08-21 GQA paged decode h32 kv8 d128 ps1gqa-paged-attention3802026-08-21 GQA paged prefill causal h32 kv8 d128 ps1gqa-paged-attention912026-08-21 GQA paged prefill causal h32 kv4 d128 ps1gqa-paged-attention892026-08-21 GQA paged decode h20 kv4 d128 ps1gqa-paged-attention402026-03-28 GQA paged decode h20 kv4 d128 ps64gqa-paged-attention402026-03-28 GQA paged decode h5 kv1 d128 ps1gqa-paged-attention202026-04-12 GQA paged decode h24 kv4 d128 ps64gqa-paged-attention202026-03-31 GQA paged decode h24 kv8 d128 ps1gqa-paged-attention202026-04-09 GQA paged decode h32 kv4 d128 ps64gqa-paged-attention202026-04-09 GQA paged prefill causal h16 kv2 d128 ps64gqa-paged-attention202026-04-09 GQA paged prefill causal h20 kv4 d128 ps1gqa-paged-attention202026-03-28 GQA paged prefill causal h20 kv4 d128 ps64gqa-paged-attention202026-03-28 GQA paged prefill causal h24 kv4 d128 ps1gqa-paged-attention202026-03-31 GQA paged prefill causal h24 kv8 d128 ps1gqa-paged-attention202026-04-12 GQA paged prefill causal h40 kv10 d128 ps1gqa-paged-attention202026-04-07 GQA paged prefill causal h24 kv8 d128 ps64gqa-paged-attention192026-04-12 GQA paged decode h32 kv8 d128 ps64gqa-paged-attention182026-04-30 GQA paged prefill causal h5 kv1 d128 ps1gqa-paged-attention142026-04-12 GQA paged prefill causal h16 kv1 d128 ps64gqa-paged-attention142026-04-05 GQA paged decode h32 kv8 d64 ps1gqa-paged-attention82026-04-20 GQA paged decode h32 kv8 d64 ps64gqa-paged-attention82026-04-20 GQA paged prefill causal h32 kv8 d64 ps1gqa-paged-attention82026-04-20 GQA paged prefill causal h32 kv8 d64 ps64gqa-paged-attention82026-04-20 GQA paged prefill causal h32 kv4 d128 ps64gqa-paged-attention52026-04-09 GQA paged prefill causal h24 kv4 d128 ps64gqa-paged-attention42026-03-31 GQA paged decode h6 kv1 d128 ps1gqa-paged-attentionnone yet—GQA paged decode h6 kv1 d128 ps64gqa-paged-attentionnone yet—GQA paged decode h16 kv2 d128 ps1gqa-paged-attentionnone yet—GQA paged decode h16 kv2 d128 ps64gqa-paged-attentionnone yet—GQA paged decode h24 kv8 d128 ps64gqa-paged-attentionnone yet—GQA paged decode h32 kv16 d128 ps1gqa-paged-attentionnone yet—GQA paged decode h32 kv16 d128 ps64gqa-paged-attentionnone yet—GQA paged decode h48 kv8 d128 ps1gqa-paged-attentionnone yet—GQA paged prefill causal h5 kv1 d128 ps64gqa-paged-attentionnone yet—GQA paged prefill causal h6 kv1 d128 ps1gqa-paged-attentionnone yet—GQA paged prefill causal h6 kv1 d128 ps64gqa-paged-attentionnone yet—GQA paged prefill causal h16 kv2 d128 ps1gqa-paged-attentionnone yet—GQA paged prefill causal h32 kv8 d128 ps64gqa-paged-attentionnone yet—GQA paged prefill causal h32 kv16 d128 ps1gqa-paged-attentionnone yet—GQA paged prefill causal h32 kv16 d128 ps64gqa-paged-attentionnone yet—

4 folded into an equivalent definition · 15 indexed without runs · How counting works

Ranked only against runs that measured the same thing.