Sparse MoE routing and experts
Sparse Mixture-of-Experts block with top-10 routing across 512 experts. Includes router network, top-k expert selection with normalized weights, parallel expert computation with weighted aggregation, and shared expert with learned gating. This is a complete MoE forward pass with dynamic token-to-expert routing.
Reported evidence · last observed 2026-08-13. Reported by source; not independently reproduced. No public source. License unknown.
Current records
Workloadsuite of 16 cases · mean latencyhidden_size = 2048 · num_experts = 512 · num_experts_per_tok = 10 · moe_intermediate_size = 512 · shared_expert_intermediate_size = 51216 cases
Not measured on H100 for this workload. Challenges →
1747.6µs86% SOL1.00×Reported · license unknown · no source2026-08-13
Measured exactly what you asked. Reported by source; not independently reproduced.
2748.2µs86% SOL1.00×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
3759.2µs85% SOL1.02×Reported · license unknown · no source2026-07-18
Measured exactly what you asked. Reported by source; not independently reproduced.
4759.9µs85% SOL1.02×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
5852.3µs82% SOL1.14×Reported · license unknown · no source2026-07-22
Measured exactly what you asked. Reported by source; not independently reproduced.
6872.6µs80% SOL1.17×Reported · license unknown · no source2026-08-18
Measured exactly what you asked. Reported by source; not independently reproduced.
7947.4µs78% SOL1.27×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
81.08ms74% SOL1.44×Reported · license unknown · no source2026-08-17
Measured exactly what you asked. Reported by source; not independently reproduced.
91.08ms73% SOL1.45×Reported · license unknown · no source2026-07-25
Measured exactly what you asked. Reported by source; not independently reproduced.
101.08ms74% SOL1.45×Reported · license unknown · no source2026-08-18
Measured exactly what you asked. Reported by source; not independently reproduced.
111.14ms72% SOL1.53×Reported · license unknown · no source2026-07-13
Measured exactly what you asked. Reported by source; not independently reproduced.
121.19ms70% SOL1.59×Reported · license unknown · no source2026-08-08
Measured exactly what you asked. Reported by source; not independently reproduced.
131.30ms67% SOL1.74×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.
141.73ms57% SOL2.31×Reported · license unknown · no source2026-07-23
Measured exactly what you asked. Reported by source; not independently reproduced.
152.23ms49% SOL2.98×Reported · license unknown · no source2026-07-10
Measured exactly what you asked. Reported by source; not independently reproduced.