qwen3-30b-a3b
0 of 18 operations have a usable best known on NVIDIA B200 · 18 without one
Best known on NVIDIA B200
GQA paged decode h32 kv4 d128 ps1bf16 · [1, 32, 128] · num pages 17 · num kv indices 2flashinfer / wrapper78fd04not usable8.23µsReported · Apache-2.0 · source2025-10-21stale
No entry in this cohort passes the usability policy (no install recipe); the fastest known is shown. Measured by the source's own harness. Ranked by its metric under ranking-v1. 6 more entries in this cohort.
GQA paged decode h32 kv4 d128 ps64bf16 · [1, 32, 128] · num pages 122 · num kv indices 58flashinfer / wrapperda7954not usable19.5µsReported · Apache-2.0 · source2026-04-09
No entry in this cohort passes the usability policy (no install recipe); the fastest known is shown. Measured by the source's own harness. Ranked by its metric under ranking-v1.
GQA paged prefill causal h32 kv4 d128 ps1bf16 · [1, 32, 128]claude-opus-4-1 / triton07ad16not usable100.3µsReported · Apache-2.0 · source2025-10-21stale
No entry in this cohort passes the usability policy (no install recipe); the fastest known is shown. Measured by the source's own harness. Ranked by its metric under ranking-v1. 5 more entries in this cohort.
GQA paged prefill causal h32 kv4 d128 ps64bf16 · [992, 32, 128]flashinfer / wrapper50ca11not usable64.5µsReproduction-ready · Apache-2.0 · source2026-04-09
No entry in this cohort passes the usability policy (no install recipe); the fastest known is shown. Measured by the source's own harness. Ranked by its metric under ranking-v1.
GEMM n128 k2048fp16 · [34, 2048]torch_matmul_317103not usable9.45µsReported · Apache-2.0 · source2025-10-20stale
No entry in this cohort passes the usability policy (no install recipe); the fastest known is shown. Measured by the source's own harness. Ranked by its metric under ranking-v1. 6 more entries in this cohort.
GEMM n2048 k4096fp16 · [17, 4096]torch / matmul926adcnot usable13.8µsReported · Apache-2.0 · source2025-10-20stale
No entry in this cohort passes the usability policy (no install recipe); the fastest known is shown. Measured by the source's own harness. Ranked by its metric under ranking-v1. 6 more entries in this cohort.
GEMM n5120 k2048fp16 · [6, 2048]torch / matmul075b0dnot usable12.0µsReported · Apache-2.0 · source2025-10-20stale
No entry in this cohort passes the usability policy (no install recipe); the fastest known is shown. Measured by the source's own harness. Ranked by its metric under ranking-v1. 5 more entries in this cohort.
Fused add RMSNorm h2048bf16 · [2048] · batch 79gpt-o3 / triton2e18c2not usable6.19µsReported · Apache-2.0 · source2025-10-16stale
No entry in this cohort passes the usability policy (no install recipe); the fastest known is shown. Measured by the source's own harness. Ranked by its metric under ranking-v1. 7 more entries in this cohort.
RMSNorm h128bf16 · [128] · batch 24gpt-5 / cuda1ad912not usable6.17µsReported · Apache-2.0 · source2025-10-16stale
No entry in this cohort passes the usability policy (no install recipe); the fastest known is shown. Measured by the source's own harness. Ranked by its metric under ranking-v1. 8 more entries in this cohort.
RMSNorm h2048bf16 · [2048] · batch 34gemini-2.5-pro / tritonuc5ufinot usable6.17µsReported · Apache-2.0 · source2025-10-16stale
No entry in this cohort passes the usability policy (no install recipe); the fastest known is shown. Measured by the source's own harness. Ranked by its metric under ranking-v1. 6 more entries in this cohort.
GQA ragged prefill causal h32 kv4 d128bf16 · [34, 4, 128] · #641e77flashinfer / wrapperacea60not usable10.3µsReported · Apache-2.0 · source2025-10-21stale
No entry in this cohort passes the usability policy (no install recipe); the fastest known is shown. Measured by the source's own harness. Ranked by its metric under ranking-v1. 19 more entries in this cohort.
Coverage gaps on NVIDIA B200
No eligible evidence yet for these operations. Challenges →
Related tags: qwen3-30b-a3b-instruct-2507