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

⏎

gpu B200✕

20 operations match

Pick one to compare its implementations; your filters carry over

Operation

Family

Published runs

Last run

FP8 MoE expert linear33 on B200 · best 93.6 µsmoe332026-08-23 Grouped topk MoE routing backward32 on B200 · best 23.0 µsmoe322026-08-23 MoE expert batched execution with capacity factor24 on B200 · best 2.14 msmoe242026-08-19 MoE expert computation27 on B200 · best 3.67 msmoe272026-08-22 MoE expert computation with weighted accumulation23 on B200 · best 373.4 µsmoe232026-08-20 MoE expert inference batched dispatch25 on B200 · best 1.21 msmoe252026-08-21 MoE expert load balancing and token capacity backward22 on B200 · best 7.21 µsmoe222026-07-23 MoE expert MLP with load balancing28 on B200 · best 713.5 µsmoe282026-08-21 MoE expert parallel execution24 on B200 · best 2.37 msmoe242026-08-21 MoE expert parallel execution backward34 on B200 · best 12.6 msmoe342026-08-21 MoE expert parallel execution with weighted aggregation22 on B200 · best 4.93 msmoe222026-08-20 MoE expert SwiGLU with down projection backward48 on B200 · best 42.2 µsmoe482026-08-24 MoE expert token radix sort with prefix sum30 on B200 · best 5.13 µsmoe302026-08-24 MoE group score aggregation and masking19 on B200 · best 3.51 µsmoe192026-07-23 MoE layer complete forward with residual30 on B200 · best 1.00 msmoe302026-08-23 MoE sparse routing and dispatch30 on B200 · best 5.74 msmoe302026-08-16 MoE sparse routing and dispatch31 on B200 · best 218.3 µsmoe312026-08-20 MoE sparse routing and dispatch backward22 on B200 · best 15.2 µsmoe222026-08-23 MoE training token repeat and expert computation23 on B200 · best 4.47 msmoe232026-08-23 Sparse MoE routing and experts22 on B200 · best 740.0 µsmoe222026-08-18

Ranked only against runs that measured the same thing.