deepseek-v3-1-terminus
best known per operation on the selected GPU
2 operations · 2 families · 85 eligible runs
0 of 2 operations have a usable best known on NVIDIA B200 · 2 without one
Best known on NVIDIA B200
Operation · workload
Best known
Record
Trust
Observed
mla-paged-attention1 operation
FP8 MLA KV compression projectionsuite of 16 casesRIAC as wellnot usable162.4µsReported · license unknown · no source2026-08-22
No entry in this cohort passes the usability policy (no public source, no install recipe, license unknown); the fastest known is shown. Measured by the source's own harness. Ranked by its metric under ranking-v1. 38 more entries in this cohort.
no sourcelicense unknownno install recipeCohort on the operation page →Run detail →
rope1 operation
FP8 YaRN RoPE embeddingsuite of 16 casesdoubleAInot usable2.00µsReported · license unknown · no source2026-07-10
No entry in this cohort passes the usability policy (no public source, no install recipe, license unknown); the fastest known is shown. Measured by the source's own harness. Ranked by its metric under ranking-v1. 29 more entries in this cohort.
no sourcelicense unknownno install recipeCohort on the operation page →Run detail →
Coverage gaps on NVIDIA B200
Operation
Family
Status
mla-paged-attention
measured · no usable entry (no public source, no install recipe, license unknown)
rope
measured · no usable entry (no public source, no install recipe, license unknown)
No eligible evidence yet for these operations. Challenges →
Related tags: deepseek-v3
Sources: NVIDIA SOL-ExecBench (2026-08-24)How comparison worksJSON