Amir M. Mir | SF Tensor
Competition author · sol-execbench/user/Amir M. Mir | SF Tensor · license unknown
Records held
61as of 2026-10-01 19:24 UTC
Entries
1,381
Runs
1,381
Is this you? Claim this project ›
A claim grants attribution and metadata maintenance, never the right to edit evidence. Evidence claims are reviewed by a maintainer.
Current records
Operation · workload
History
Record
Implementation
Hardware
Since
GQA ragged prefill causal h32 kv4 d128suite of 15 cases
7.00µs
4.2% faster · was 7.31 µs
B200
2026-08-21
NVFP4 grouped query attention with KV repeatsuite of 16 cases
290.0µs
2.4% faster · was 297.2 µs
B200
2026-08-20
NVFP4 fused decoder layer pre norm attentionsuite of 16 cases
589.1µs
7.8% faster · was 639.2 µs
B200
2026-08-20
Show all 40 records ›Showing all 40 records ⌄
FP8 vision attention output projectionsuite of 16 cases
20.0µs
0.1% faster · was 20.0 µs
B200
2026-07-04
FP8 vision attention output projectionsuite of 16 cases
17.3µs
1.6% faster · was 17.6 µs
B200
2026-08-20
Flux timestep guidance projection embeddingsuite of 16 cases
27.1µs
16.5% faster · was 32.5 µs
B200
2026-08-19
Patch embed to joint attention inputsuite of 16 cases
254.6µs
1.8% faster · was 259.3 µs
B200
2026-08-21
Altup predict correction cycle backwardsuite of 16 cases
56.4µs
20.9% faster · was 71.3 µs
B200
2026-08-19
Dual stream attention with conditional cross attentionsuite of 16 cases
305.2µs
11.6% faster · was 345.3 µs
B200
2026-08-22
MoE expert parallel execution with weighted aggregationsuite of 16 cases
5.41ms
1.3% faster · was 5.48 ms
B200
2026-07-06
Expert weighted aggregation with shared expertsuite of 16 cases
714.4µs
2.0% faster · was 728.9 µs
B200
2026-08-16
Audio encoder varlen attention with chunking backwardsuite of 16 cases
64.9µs
9.4% faster · was 71.7 µs
B200
2026-08-22
MoE expert computation with weighted accumulationsuite of 16 cases
373.4µs
2.4% faster · was 382.6 µs
B200
2026-08-17
Fused attention QK matmul scale mask softmax backwardsuite of 16 cases
137.0µs
1.1% faster · was 138.5 µs
B200
2026-08-22
Altup hidden state collapse with magnitude normalizationsuite of 16 cases
96.8µs
6.4% faster · was 103.5 µs
B200
2026-07-10
Flux ada layer norm zero modulation extractionsuite of 16 cases
68.9µs
3.3% faster · was 71.3 µs
B200
2026-08-24
GQA paged prefill causal h32 kv4 d128 ps1suite of 30 cases
11.2µs
17.7% faster · was 13.6 µs
B200
2026-07-04
GQA paged prefill causal h32 kv4 d128 ps1suite of 30 cases
5.31µs
2.2% faster · was 5.43 µs
B200
2026-08-17
NVFP4 attention QKV projection with RoPEsuite of 16 cases
86.1µs
19.3% faster · was 106.7 µs
B200
2026-08-01
NVFP4 MoE routing with topk selectionsuite of 16 cases
22.4µs
33.2% faster · was 33.5 µs
B200
2026-07-22
21 more records · Open in the records ledger →
Measured entries
Implementation
Operation
Runtime
Best latency
Evidence
Availability
unknown
1.28µs
Reported
license unknown · no source
unknown
1.67µs
Reported
license unknown · no source
unknown
1.76µs
Reported
license unknown · no source
unknown
1.81µs
Reported
license unknown · no source
unknown
1.82µs
Reported
license unknown · no source
unknown
2.01µs
Reported
license unknown · no source
unknown
2.04µs
Reported
license unknown · no source
unknown
2.32µs
Reported
license unknown · no source
unknown
2.38µs
Reported
license unknown · no source
Show all 40 entries ›Showing all 40 entries ⌄
unknown
2.49µs
Reported
license unknown · no source
unknown
2.56µs
Reported
license unknown · no source
unknown
2.63µs
Reported
license unknown · no source
unknown
2.64µs
Reported
license unknown · no source
unknown
2.69µs
Reported
license unknown · no source
unknown
2.74µs
Reported
license unknown · no source
unknown
2.91µs
Reported
license unknown · no source
unknown
3.09µs
Reported
license unknown · no source
unknown
3.12µs
Reported
license unknown · no source
unknown
3.73µs
Reported
license unknown · no source
unknown
3.77µs
Reported
license unknown · no source
unknown
3.77µs
Reported
license unknown · no source
Showing 40 of 1381; the full list is in the API dossier below.
Record activity
Nov
Dec
Jan
Feb
Mar
10Apr
8May
14Jun
211Jul
70Aug
Sep
Oct
Sources: NVIDIA SOL-ExecBench (2026-08-24)last observed 2026-08-24JSON