Skip to content
KernelIndex
Search⌘K

NVFP4 group GEMM

455 eligible runs
gemm

NVFP4 block-scaled group matrix multiply for B200: g independent GEMMs, each with its own (m, n, k) problem size, over nvfp4 e2m1 operands with fp8 scale factors per 16 elements of k and fp16 outputs. Per-group m and n are divisible by the kernel's mma tile, k by 256, and the batch dim l is always 1. Tensor shapes below are per group.

Fastest reported
13.1µsgeomean
submission 505727gau.nernst · License unknown · python

Reported evidence · last observed 2026-02-21. Reported by source; not independently reproduced. License unknown.

Current records

Not measured on H100 for this workload. Challenges →

Source-native comparison · GPU NVIDIA B200 · Workload suite of 4 cases · geomean score · Protocol reference-kernels eval.py · geomean · 310 results · last observed 2026-02-21Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
13.1µs
1.00×
Reported · license unknown · source
2026-02-21stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
2
13.2µs
1.01×
Reported · license unknown · source
2026-02-20stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
3
submission 506188nataliakokoromyti
13.8µs
1.05×
Reported · license unknown · source
2026-02-21stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
4
13.8µs
1.06×
Reported · license unknown · source
2026-02-21stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
5
13.8µs
1.06×
Reported · license unknown · source
2026-02-21stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
6
13.9µs
1.06×
Reported · license unknown · source
2026-02-21stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
7
13.9µs
1.06×
Reported · license unknown · source
2026-02-21stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
8
14.1µs
1.08×
Reported · license unknown · source
2026-02-21stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
9
14.1µs
1.08×
Reported · license unknown · source
2026-02-21stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
10
14.2µs
1.09×
Reported · license unknown · source
2026-02-20stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
11
14.5µs
1.11×
Reported · license unknown · source
2026-02-18stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
12
14.8µs
1.13×
Reported · license unknown · source
2026-02-12stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
13
14.9µs
1.14×
Reported · license unknown · source
2026-02-09stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
14
15.0µs
1.14×
Reported · license unknown · source
2026-02-09stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
15
15.1µs
1.15×
Reported · license unknown · source
2026-02-18stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
16
15.1µs
1.15×
Reported · license unknown · source
2026-02-20stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
17
15.1µs
1.16×
Reported · license unknown · source
2026-02-19stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
18
15.2µs
1.16×
Reported · license unknown · source
2026-02-20stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
19
15.2µs
1.16×
Reported · license unknown · source
2026-02-10stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
20
15.2µs
1.16×
Reported · license unknown · source
2026-02-20stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
21
15.2µs
1.16×
Reported · license unknown · source
2026-02-21stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
22
15.3µs
1.17×
Reported · license unknown · source
2026-02-20stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
23
15.4µs
1.18×
Reported · license unknown · source
2026-02-20stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
24
15.4µs
1.18×
Reported · license unknown · source
2026-02-20stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
25
15.4µs
1.18×
Reported · license unknown · source
2026-02-18stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
26
15.5µs
1.18×
Reported · license unknown · source
2026-02-15stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
27
15.5µs
1.18×
Reported · license unknown · source
2026-02-17stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
28
15.7µs
1.20×
Reported · license unknown · source
2026-02-20stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
29
15.7µs
1.20×
Reported · license unknown · source
2026-02-17stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
30
15.8µs
1.20×
Reported · license unknown · source
2026-02-14stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
31
15.8µs
1.21×
Reported · license unknown · no source
2026-02-17stale

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
32
15.9µs
1.22×
Reported · license unknown · source
2026-02-20stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
33
15.9µs
1.22×
Reported · license unknown · source
2026-02-16stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
34
16.0µs
1.22×
Reported · license unknown · source
2026-02-16stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
35
16.0µs
1.22×
Reported · license unknown · source
2026-02-12stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
36
submission 506083nataliakokoromyti
16.0µs
1.22×
Reported · license unknown · source
2026-02-21stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
37
16.1µs
1.23×
Reported · license unknown · source
2026-02-14stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
38
16.1µs
1.23×
Reported · license unknown · source
2026-02-20stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
39
16.1µs
1.23×
Reported · license unknown · source
2026-02-14stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
40
16.2µs
1.24×
Reported · license unknown · source
2026-02-14stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
41
16.4µs
1.25×
Reported · license unknown · source
2026-02-14stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
42
submission 505064nataliakokoromyti
16.6µs
1.27×
Reported · license unknown · source
2026-02-20stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
43
16.7µs
1.27×
Reported · license unknown · source
2026-02-10stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
44
16.8µs
1.28×
Reported · license unknown · source
2026-02-14stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
45
16.9µs
1.29×
Reported · license unknown · source
2026-02-09stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
46
17.1µs
1.31×
Reported · license unknown · source
2026-02-08stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
47
17.3µs
1.32×
Reported · license unknown · source
2026-02-07stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
48
17.6µs
1.34×
Reported · license unknown · source
2026-02-13stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
49
17.6µs
1.35×
Reported · license unknown · source
2026-02-10stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
50
17.7µs
1.35×
Reported · license unknown · source
2026-02-09stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
51
submission 498506Naturalseeker
17.8µs
1.36×
Reported · license unknown · source
2026-02-18stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
52
17.8µs
1.36×
Reported · license unknown · source
2026-02-09stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
53
submission 498323Naturalseeker
17.8µs
1.36×
Reported · license unknown · source
2026-02-18stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
54
17.9µs
1.37×
Reported · license unknown · source
2026-02-09stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
55
17.9µs
1.37×
Reported · license unknown · source
2026-02-09stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
56
17.9µs
1.37×
Reported · license unknown · source
2026-02-08stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
57
submission 495386Naturalseeker
17.9µs
1.37×
Reported · license unknown · source
2026-02-17stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
58
17.9µs
1.37×
Reported · license unknown · source
2026-02-08stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
59
18.1µs
1.38×
Reported · license unknown · source
2026-02-08stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →
60
18.2µs
1.39×
Reported · license unknown · source
2026-02-07stale

Measured exactly what you asked. Reported by source; not independently reproduced.

source mirroredlicense unknownno install recipeView source →Run detail →

250 more rows in this comparison group. Open all in search →

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
submission 506336nataliakokoromyti
python
11.4µs
1.00×
Reported
license unknown · source
python
16.0µs
1.40×
Reported
license unknown · source
python
18.1µs
1.59×
Reported
license unknown · source
python
18.4µs
1.61×
Reported
license unknown · source
submission 500890mufeez-amjad
python
21.0µs
1.84×
Reported
license unknown · source
python
21.4µs
1.87×
Reported
license unknown · source
python
21.6µs
1.89×
Reported
license unknown · source
python
26.4µs
2.31×
Reported
license unknown · source
Show all 455 implementations ›
python
27.2µs
2.38×
Reported
license unknown · source
python
30.3µs
2.65×
Reported
license unknown · source
python
30.8µs
2.70×
Reported
license unknown · source
python
31.2µs
2.73×
Reported
license unknown · source
python
32.1µs
2.81×
Reported
license unknown · source
python
32.3µs
2.83×
Reported
license unknown · source
python
34.1µs
2.98×
Reported
license unknown · source
python
37.8µs
3.31×
Reported
license unknown · source
python
37.9µs
3.32×
Reported
license unknown · source
python
38.5µs
3.37×
Reported
license unknown · source
submission 442449Arseni Ivanov
python
45.7µs
4.00×
Reported
license unknown · source
python
46.6µs
4.08×
Reported
license unknown · source
python
47.8µs
4.18×
Reported
license unknown · source
submission 487985nataliakokoromyti
python
75.0µs
6.57×
Reported
license unknown · source
submission 500064nataliakokoromyti
python
35.9µs
3.14×
Reported
license unknown · source
submission 503565nataliakokoromyti
python
24.3µs
2.13×
Reported
license unknown · source
python
191.2µs
16.74×
Reported
license unknown · source
python
247.6µs
21.67×
Reported
license unknown · source
python
242.9µs
21.26×
Reported
license unknown · source
python
240.7µs
21.07×
Reported
license unknown · source
python
237.7µs
20.81×
Reported
license unknown · source
python
236.1µs
20.67×
Reported
license unknown · source
python
233.4µs
20.42×
Reported
license unknown · source
submission 487322mufeez-amjad
python
32.1µs
2.81×
Reported
license unknown · source
submission 490526mufeez-amjad
python
23.7µs
2.08×
Reported
license unknown · source
python
35.6µs
3.11×
Reported
license unknown · source
python
35.2µs
3.08×
Reported
license unknown · source
python
34.3µs
3.00×
Reported
license unknown · source
python
33.5µs
2.94×
Reported
license unknown · source
python
31.9µs
2.79×
Reported
license unknown · source
python
22.0µs
1.92×
Reported
license unknown · source
python
21.4µs
1.88×
Reported
license unknown · source
python
21.4µs
1.88×
Reported
license unknown · source
python
21.4µs
1.88×
Reported
license unknown · source
python
202.8µs
17.75×
Reported
license unknown · source
python
164.7µs
14.42×
Reported
license unknown · source
python
52.0µs
4.56×
Reported
license unknown · source
python
33.4µs
2.92×
Reported
license unknown · source
python
23.8µs
2.08×
Reported
license unknown · source
python
23.7µs
2.08×
Reported
license unknown · source
python
45.3µs
3.96×
Reported
license unknown · source
python
42.5µs
3.72×
Reported
license unknown · source
python
29.7µs
2.60×
Reported
license unknown · source
python
26.5µs
2.32×
Reported
license unknown · source
python
38.9µs
3.40×
Reported
license unknown · source
python
28.4µs
2.48×
Reported
license unknown · source
python
27.9µs
2.44×
Reported
license unknown · source
python
27.5µs
2.41×
Reported
license unknown · source
python
27.2µs
2.38×
Reported
license unknown · source
python
216.3µs
18.93×
Reported
license unknown · source
python
68.8µs
6.02×
Reported
license unknown · source
python
53.8µs
4.71×
Reported
license unknown · source
python
274.5µs
24.02×
Reported
license unknown · source
python
94.3µs
8.25×
Reported
license unknown · source
python
69.3µs
6.07×
Reported
license unknown · source
python
45.7µs
4.00×
Reported
license unknown · source
python
48.7µs
4.26×
Reported
license unknown · source
python
37.8µs
3.30×
Reported
license unknown · source
python
90.4µs
7.91×
Reported
license unknown · source
python
42.8µs
3.74×
Reported
license unknown · source
python
1.03ms
89.90×
Reported
license unknown · source
python
502.8µs
44.01×
Reported
license unknown · source
python
96.5µs
8.45×
Reported
license unknown · source
python
85.8µs
7.51×
Reported
license unknown · source
python
60.6µs
5.30×
Reported
license unknown · source
python
60.2µs
5.27×
Reported
license unknown · source
python
60.1µs
5.26×
Reported
license unknown · source
python
59.5µs
5.21×
Reported
license unknown · source
python
56.5µs
4.94×
Reported
license unknown · source
python
56.3µs
4.93×
Reported
license unknown · source
python
56.3µs
4.93×
Reported
license unknown · source
python
57.3µs
5.02×
Reported
license unknown · source
python
57.2µs
5.01×
Reported
license unknown · source
python
57.2µs
5.01×
Reported
license unknown · source
python
57.2µs
5.01×
Reported
license unknown · source
python
57.2µs
5.00×
Reported
license unknown · source
python
43.6µs
3.82×
Reported
license unknown · source
python
39.8µs
3.48×
Reported
license unknown · source
python
38.5µs
3.37×
Reported
license unknown · source
python
1.82ms
159.03×
Reported
license unknown · source
python
264.8µs
23.17×
Reported
license unknown · source
python
246.8µs
21.60×
Reported
license unknown · source
python
113.5µs
9.93×
Reported
license unknown · source
python
94.5µs
8.27×
Reported
license unknown · source
python
79.4µs
6.95×
Reported
license unknown · source
python
75.1µs
6.58×
Reported
license unknown · source
python
61.4µs
5.37×
Reported
license unknown · source
python
57.5µs
5.03×
Reported
license unknown · source
python
57.1µs
5.00×
Reported
license unknown · source
python
52.2µs
4.57×
Reported
license unknown · source
python
48.3µs
4.22×
Reported
license unknown · source
python
47.1µs
4.12×
Reported
license unknown · source
python
46.9µs
4.11×
Reported
license unknown · source
python
13.1µs
1.15×
Reported
license unknown · source
submission 506188nataliakokoromyti
python
13.8µs
1.21×
Reported
license unknown · source
python
13.8µs
1.21×
Reported
license unknown · source
python
14.1µs
1.24×
Reported
license unknown · source
python
14.8µs
1.30×
Reported
license unknown · source
python
15.1µs
1.32×
Reported
license unknown · source
python
15.2µs
1.33×
Reported
license unknown · source
python
15.4µs
1.35×
Reported
license unknown · source
python
15.9µs
1.39×
Reported
license unknown · source
submission 498506Naturalseeker
python
17.8µs
1.56×
Reported
license unknown · source
python
17.9µs
1.56×
Reported
license unknown · source
python
18.7µs
1.64×
Reported
license unknown · source
python
19.6µs
1.71×
Reported
license unknown · source
python
19.6µs
1.72×
Reported
license unknown · source
python
20.3µs
1.77×
Reported
license unknown · source
python
20.7µs
1.81×
Reported
license unknown · source
python
20.8µs
1.82×
Reported
license unknown · source
submission 505548mufeez-amjad
python
21.0µs
1.84×
Reported
license unknown · source
python
21.1µs
1.84×
Reported
license unknown · source
python
22.4µs
1.96×
Reported
license unknown · source
python
22.4µs
1.96×
Reported
license unknown · source
python
22.5µs
1.97×
Reported
license unknown · source
submission 484166Arseni Ivanov
python
22.7µs
1.99×
Reported
license unknown · source
python
22.8µs
1.99×
Reported
license unknown · source
python
23.3µs
2.04×
Reported
license unknown · source
python
23.5µs
2.06×
Reported
license unknown · source
python
23.9µs
2.09×
Reported
license unknown · source
python
25.2µs
2.20×
Reported
license unknown · source
python
25.4µs
2.22×
Reported
license unknown · source
python
26.0µs
2.28×
Reported
license unknown · source
python
26.3µs
2.30×
Reported
license unknown · source
python
27.2µs
2.38×
Reported
license unknown · source
submission 493184rylanmalarchick
python
28.1µs
2.46×
Reported
license unknown · source
python
28.5µs
2.49×
Reported
license unknown · source
python
28.8µs
2.52×
Reported
license unknown · source
python
29.1µs
2.54×
Reported
license unknown · source
submission 505354Jiannan Wang
python
29.2µs
2.55×
Reported
license unknown · source
python
32.7µs
2.86×
Reported
license unknown · source
submission 488633yolojewjitsu
python
32.9µs
2.88×
Reported
license unknown · source
python
34.3µs
3.00×
Reported
license unknown · source
python
38.4µs
3.36×
Reported
license unknown · source
python
41.5µs
3.64×
Reported
license unknown · source
python
53.1µs
4.65×
Reported
license unknown · source
python
53.2µs
4.65×
Reported
license unknown · source
submission 438751Sheheryar Ahmad
python
55.3µs
4.84×
Reported
license unknown · source
submission 483123farkhanda ghaffar
python
55.4µs
4.85×
Reported
license unknown · source
submission 475463Jitesh Jain US
python
55.4µs
4.85×
Reported
license unknown · source
python
61.4µs
5.37×
Reported
license unknown · source
python
62.8µs
5.50×
Reported
license unknown · source
python
78.1µs
6.84×
Reported
license unknown · source
python
97.0µs
8.49×
Reported
license unknown · source
python
97.7µs
8.55×
Reported
license unknown · source
python
102.5µs
8.97×
Reported
license unknown · source
python
125.7µs
11.00×
Reported
license unknown · source
submission 476068kaiming-cheng
python
138.1µs
12.09×
Reported
license unknown · source
python
193.0µs
16.89×
Reported
license unknown · source
python
199.2µs
17.43×
Reported
license unknown · source
submission 473911Clark Kitchen
python
200.8µs
17.57×
Reported
license unknown · source
submission 477850lucifer_0000007
python
230.2µs
20.14×
Reported
license unknown · source
python
236.9µs
20.73×
Reported
license unknown · source
python
238.2µs
20.85×
Reported
license unknown · source
python
240.3µs
21.03×
Reported
license unknown · source
python
240.7µs
21.06×
Reported
license unknown · source
python
242.6µs
21.24×
Reported
license unknown · source
python
243.7µs
21.33×
Reported
license unknown · source
python
244.8µs
21.42×
Reported
license unknown · source
python
294.3µs
25.76×
Reported
license unknown · source
python
299.7µs
26.23×
Reported
license unknown · source
submission 494881我爱拆拆
python
355.3µs
31.10×
Reported
license unknown · source
python
8.68ms
759.35×
Reported
license unknown · source
python
9.19ms
804.46×
Reported
license unknown · source
python
184.0µs
16.11×
Reported
license unknown · source
python
30.6µs
2.67×
Reported
license unknown · source
python
30.0µs
2.63×
Reported
license unknown · source
python
18.2µs
1.59×
Reported
license unknown · source
python
16.0µs
1.40×
Reported
license unknown · source
python
15.1µs
1.32×
Reported
license unknown · source
python
13.2µs
1.16×
Reported
license unknown · source
submission 399698nataliakokoromyti
python
240.5µs
21.05×
Reported
license unknown · source
submission 399793nataliakokoromyti
python
219.7µs
19.23×
Reported
license unknown · source
submission 399847nataliakokoromyti
python
218.5µs
19.13×
Reported
license unknown · source
submission 400223nataliakokoromyti
python
79.9µs
7.00×
Reported
license unknown · source
submission 492509nataliakokoromyti
python
55.4µs
4.85×
Reported
license unknown · source
submission 498371nataliakokoromyti
python
34.8µs
3.04×
Reported
license unknown · source
submission 498449nataliakokoromyti
python
34.6µs
3.03×
Reported
license unknown · source
submission 498453nataliakokoromyti
python
30.4µs
2.66×
Reported
license unknown · source
submission 503489nataliakokoromyti
python
26.7µs
2.34×
Reported
license unknown · source
submission 503497nataliakokoromyti
python
26.3µs
2.30×
Reported
license unknown · source
submission 504695nataliakokoromyti
python
24.8µs
2.17×
Reported
license unknown · source
submission 504852nataliakokoromyti
python
20.2µs
1.77×
Reported
license unknown · source
submission 505064nataliakokoromyti
python
16.6µs
1.45×
Reported
license unknown · source
submission 506083nataliakokoromyti
python
16.0µs
1.40×
Reported
license unknown · source
python
19.7µs
1.73×
Reported
license unknown · source
python
15.7µs
1.37×
Reported
license unknown · source
python
15.3µs
1.34×
Reported
license unknown · source
python
13.9µs
1.22×
Reported
license unknown · source
python
13.9µs
1.21×
Reported
license unknown · source
python
13.8µs
1.21×
Reported
license unknown · source
python
39.1µs
3.42×
Reported
license unknown · source
python
32.8µs
2.87×
Reported
license unknown · source
python
24.3µs
2.12×
Reported
license unknown · source
python
23.9µs
2.09×
Reported
license unknown · source
python
21.2µs
1.85×
Reported
license unknown · source
python
20.0µs
1.75×
Reported
license unknown · source
python
18.4µs
1.61×
Reported
license unknown · source
python
18.4µs
1.61×
Reported
license unknown · source
python
17.3µs
1.52×
Reported
license unknown · source
python
15.2µs
1.33×
Reported
license unknown · source
python
15.1µs
1.32×
Reported
license unknown · source
python
14.5µs
1.27×
Reported
license unknown · source
python
14.2µs
1.24×
Reported
license unknown · source
python
14.1µs
1.24×
Reported
license unknown · source
python
15.0µs
1.31×
Reported
license unknown · source
python
14.9µs
1.31×
Reported
license unknown · source
python
16.8µs
1.47×
Reported
license unknown · source
python
15.2µs
1.33×
Reported
license unknown · source
python
15.2µs
1.33×
Reported
license unknown · source
python
108.9µs
9.53×
Reported
license unknown · source
python
37.0µs
3.23×
Reported
license unknown · source
python
27.1µs
2.37×
Reported
license unknown · source
python
24.9µs
2.18×
Reported
license unknown · source
python
21.6µs
1.89×
Reported
license unknown · source
python
16.9µs
1.47×
Reported
license unknown · source
python
16.7µs
1.46×
Reported
license unknown · source
python
16.1µs
1.41×
Reported
license unknown · source
python
16.0µs
1.40×
Reported
license unknown · source
python
15.9µs
1.39×
Reported
license unknown · source
python
15.7µs
1.38×
Reported
license unknown · source
python
15.5µs
1.36×
Reported
license unknown · source
python
15.4µs
1.35×
Reported
license unknown · source
python
15.4µs
1.35×
Reported
license unknown · source
python
33.3µs
2.91×
Reported
license unknown · source
python
21.9µs
1.92×
Reported
license unknown · source
python
19.4µs
1.69×
Reported
license unknown · source
python
19.0µs
1.66×
Reported
license unknown · source
python
18.8µs
1.64×
Reported
license unknown · source
python
18.8µs
1.64×
Reported
license unknown · source
python
18.5µs
1.62×
Reported
license unknown · source
python
18.3µs
1.61×
Reported
license unknown · source
python
18.3µs
1.60×
Reported
license unknown · source
python
17.9µs
1.57×
Reported
license unknown · source
python
17.1µs
1.50×
Reported
license unknown · source
python
16.1µs
1.41×
Reported
license unknown · source
python
15.8µs
1.38×
Reported
license unknown · source
python
15.5µs
1.36×
Reported
license unknown · source
python
21.2µs
1.86×
Reported
license unknown · source
python
19.3µs
1.69×
Reported
license unknown · source
python
18.5µs
1.62×
Reported
license unknown · source
python
17.8µs
1.56×
Reported
license unknown · source
python
17.7µs
1.55×
Reported
license unknown · source
python
17.6µs
1.54×
Reported
license unknown · source
python
17.6µs
1.54×
Reported
license unknown · source
python
16.4µs
1.43×
Reported
license unknown · source
python
16.2µs
1.42×
Reported
license unknown · source
python
16.1µs
1.41×
Reported
license unknown · source
submission 495386Naturalseeker
python
17.9µs
1.57×
Reported
license unknown · source
submission 498323Naturalseeker
python
17.8µs
1.56×
Reported
license unknown · source
python
43.6µs
3.81×
Reported
license unknown · source
python
41.3µs
3.61×
Reported
license unknown · source
python
31.1µs
2.72×
Reported
license unknown · source
python
23.7µs
2.07×
Reported
license unknown · source
python
21.5µs
1.88×
Reported
license unknown · source
python
20.4µs
1.78×
Reported
license unknown · source
python
18.5µs
1.62×
Reported
license unknown · source
python
18.5µs
1.62×
Reported
license unknown · source
python
18.4µs
1.61×
Reported
license unknown · source
python
18.3µs
1.60×
Reported
license unknown · source
python
18.2µs
1.59×
Reported
license unknown · source
python
18.1µs
1.59×
Reported
license unknown · source
python
17.9µs
1.57×
Reported
license unknown · source
python
17.9µs
1.56×
Reported
license unknown · source
python
20.7µs
1.81×
Reported
license unknown · source
python
247.0µs
21.62×
Reported
license unknown · source
python
115.9µs
10.14×
Reported
license unknown · source
python
107.5µs
9.41×
Reported
license unknown · source
python
91.1µs
7.97×
Reported
license unknown · source
python
61.1µs
5.35×
Reported
license unknown · source
python
37.9µs
3.32×
Reported
license unknown · source
python
37.8µs
3.31×
Reported
license unknown · source
python
21.5µs
1.88×
Reported
license unknown · source
python
21.4µs
1.88×
Reported
license unknown · source
python
21.3µs
1.86×
Reported
license unknown · source
python
20.5µs
1.80×
Reported
license unknown · source
python
20.1µs
1.76×
Reported
license unknown · source
python
19.7µs
1.72×
Reported
license unknown · source
python
23.3µs
2.04×
Reported
license unknown · source
python
21.2µs
1.86×
Reported
license unknown · source
python
21.1µs
1.85×
Reported
license unknown · source
python
20.2µs
1.77×
Reported
license unknown · source
python
20.0µs
1.75×
Reported
license unknown · source
python
19.8µs
1.73×
Reported
license unknown · source
python
225.7ms
19752.76×
Reported
license unknown · source
python
66.1µs
5.79×
Reported
license unknown · source
python
23.4µs
2.05×
Reported
license unknown · source
python
22.4µs
1.96×
Reported
license unknown · source
python
22.3µs
1.95×
Reported
license unknown · source
python
22.0µs
1.92×
Reported
license unknown · source
python
21.8µs
1.91×
Reported
license unknown · source
python
21.7µs
1.90×
Reported
license unknown · source
python
21.6µs
1.89×
Reported
license unknown · source
python
20.5µs
1.80×
Reported
license unknown · source
python
20.5µs
1.80×
Reported
license unknown · source
python
20.4µs
1.78×
Reported
license unknown · source
python
20.4µs
1.78×
Reported
license unknown · source
python
20.3µs
1.77×
Reported
license unknown · source
python
29.4µs
2.57×
Reported
license unknown · source
python
331.6µs
29.02×
Reported
license unknown · source
python
195.4µs
17.11×
Reported
license unknown · source
python
113.0µs
9.89×
Reported
license unknown · source
python
57.3µs
5.01×
Reported
license unknown · source
python
55.4µs
4.85×
Reported
license unknown · source
python
27.8µs
2.43×
Reported
license unknown · source
python
21.5µs
1.88×
Reported
license unknown · source
python
20.8µs
1.82×
Reported
license unknown · source
submission 487729mufeez-amjad
python
22.1µs
1.94×
Reported
license unknown · source
submission 499566mufeez-amjad
python
21.8µs
1.91×
Reported
license unknown · source
submission 500897mufeez-amjad
python
21.0µs
1.84×
Reported
license unknown · source
python
239.7µs
20.98×
Reported
license unknown · source
python
81.7µs
7.15×
Reported
license unknown · source
python
39.0µs
3.41×
Reported
license unknown · source
python
39.0µs
3.41×
Reported
license unknown · source
python
22.4µs
1.96×
Reported
license unknown · source
python
293.1µs
25.65×
Reported
license unknown · source
python
76.8µs
6.72×
Reported
license unknown · source
python
23.6µs
2.07×
Reported
license unknown · source
python
41.8µs
3.66×
Reported
license unknown · no source
python
56.8µs
4.97×
Reported
license unknown · no source
python
66.7µs
5.83×
Reported
license unknown · source
python
66.7µs
5.84×
Reported
license unknown · source
python
84.9µs
7.43×
Reported
license unknown · no source
python
104.9µs
9.18×
Reported
license unknown · source
python
108.7µs
9.52×
Reported
license unknown · no source
python
114.5µs
10.02×
Reported
license unknown · source
python
121.7µs
10.65×
Reported
license unknown · no source
python
123.0µs
10.77×
Reported
license unknown · no source
python
126.8µs
11.10×
Reported
license unknown · source
python
146.9µs
12.86×
Reported
license unknown · no source
python
156.7µs
13.72×
Reported
license unknown · no source
python
177.5µs
15.53×
Reported
license unknown · no source
python
183.7µs
16.08×
Reported
license unknown · source
python
193.2µs
16.91×
Reported
license unknown · no source
python
223.2µs
19.53×
Reported
license unknown · no source
python
239.3µs
20.95×
Reported
license unknown · source
python
255.5µs
22.37×
Reported
license unknown · source
python
255.7µs
22.38×
Reported
license unknown · source
python
261.5µs
22.89×
Reported
license unknown · no source
submission 400800lucifer_0000007
python
272.3µs
23.83×
Reported
license unknown · source
submission 400767lucifer_0000007
python
272.4µs
23.84×
Reported
license unknown · no source
python
272.7µs
23.87×
Reported
license unknown · source
python
276.1µs
24.17×
Reported
license unknown · no source
submission 400749lucifer_0000007
python
281.0µs
24.59×
Reported
license unknown · no source
python
326.7µs
28.60×
Reported
license unknown · source
python
336.1µs
29.42×
Reported
license unknown · source
python
341.1µs
29.85×
Reported
license unknown · source
python
343.2µs
30.04×
Reported
license unknown · source
python
346.4µs
30.32×
Reported
license unknown · source
python
347.3µs
30.40×
Reported
license unknown · no source
python
355.0µs
31.07×
Reported
license unknown · source
python
804.7µs
70.43×
Reported
license unknown · source
python
842.7µs
73.76×
Reported
license unknown · no source
python
890.9µs
77.98×
Reported
license unknown · source
submission 503712我爱拆拆
python
1.80ms
157.26×
Reported
license unknown · source
submission 407643我爱拆拆
python
1.82ms
159.45×
Reported
license unknown · no source
python
1.85ms
161.77×
Reported
license unknown · no source
python
22.1µs
1.93×
Reported
license unknown · no source
submission 424644Sheheryar Ahmad
python
1.86ms
162.95×
Reported
license unknown · source
python
1.89ms
165.28×
Reported
license unknown · no source
python
1.94ms
169.84×
Reported
license unknown · no source
python
2.05ms
179.31×
Reported
license unknown · source
python
2.10ms
183.62×
Reported
license unknown · source
python
15.8µs
1.39×
Reported
license unknown · no source
python
20.8µs
1.82×
Reported
license unknown · no source
python
23.3µs
2.04×
Reported
license unknown · no source
python
23.5µs
2.06×
Reported
license unknown · no source
python
24.0µs
2.10×
Reported
license unknown · no source
submission 483887Arseni Ivanov
python
24.4µs
2.13×
Reported
license unknown · no source
python
24.4µs
2.13×
Reported
license unknown · no source
python
24.7µs
2.16×
Reported
license unknown · no source
submission 503578nataliakokoromyti
python
24.8µs
2.17×
Reported
license unknown · no source
python
26.9µs
2.36×
Reported
license unknown · no source
python
27.3µs
2.39×
Reported
license unknown · no source
python
27.7µs
2.42×
Reported
license unknown · no source
python
28.0µs
2.45×
Reported
license unknown · no source
python
28.8µs
2.52×
Reported
license unknown · no source
python
29.1µs
2.55×
Reported
license unknown · no source
python
29.1µs
2.55×
Reported
license unknown · no source
submission 442445Arseni Ivanov
python
33.1µs
2.90×
Reported
license unknown · no source
submission 488551yolojewjitsu
python
33.4µs
2.93×
Reported
license unknown · no source
python
35.5µs
3.10×
Reported
license unknown · no source
submission 504517Jiannan Wang
python
35.8µs
3.13×
Reported
license unknown · no source
python
38.6µs
3.38×
Reported
license unknown · no source
python
39.0µs
3.41×
Reported
license unknown · no source
submission 493605nataliakokoromyti
python
39.0µs
3.41×
Reported
license unknown · no source
submission 489037rylanmalarchick
python
42.8µs
3.75×
Reported
license unknown · no source
python
44.1µs
3.86×
Reported
license unknown · no source
python
50.5µs
4.42×
Reported
license unknown · no source
submission 424828Sheheryar Ahmad
python
55.4µs
4.85×
Reported
license unknown · no source
submission 475321Jitesh Jain US
python
55.5µs
4.85×
Reported
license unknown · no source
python
60.2µs
5.27×
Reported
license unknown · no source
python
65.3µs
5.72×
Reported
license unknown · no source
python
67.1µs
5.87×
Reported
license unknown · no source
python
69.5µs
6.08×
Reported
license unknown · no source
python
70.0µs
6.13×
Reported
license unknown · no source
submission 488566rylanmalarchick
python
70.1µs
6.14×
Reported
license unknown · no source
python
80.4µs
7.04×
Reported
license unknown · no source
python
81.2µs
7.11×
Reported
license unknown · no source
python
89.4µs
7.83×
Reported
license unknown · no source
submission 403575Arseni Ivanov
python
94.1µs
8.23×
Reported
license unknown · no source
python
97.7µs
8.55×
Reported
license unknown · no source
python
108.4µs
9.49×
Reported
license unknown · no source
python
110.5µs
9.67×
Reported
license unknown · no source
python
119.1µs
10.43×
Reported
license unknown · no source
python
127.0µs
11.11×
Reported
license unknown · no source
python
142.2µs
12.44×
Reported
license unknown · no source
submission 488361yolojewjitsu
python
144.3µs
12.63×
Reported
license unknown · no source
python
153.6µs
13.44×
Reported
license unknown · no source
python
189.0µs
16.54×
Reported
license unknown · no source
python
199.6µs
17.47×
Reported
license unknown · no source
python
221.5µs
19.39×
Reported
license unknown · no source
python
222.3µs
19.45×
Reported
license unknown · no source
submission 472984Jitesh Jain US
python
232.6µs
20.36×
Reported
license unknown · no source
submission 417790Clark Kitchen
python
237.6µs
20.80×
Reported
license unknown · no source
python
238.0µs
20.83×
Reported
license unknown · no source
submission 424261Sheheryar Ahmad
python
238.5µs
20.88×
Reported
license unknown · no source
python
238.9µs
20.91×
Reported
license unknown · no source
submission 456532Jitesh Jain US
python
241.5µs
21.13×
Reported
license unknown · no source
python
241.6µs
21.15×
Reported
license unknown · no source
python
241.9µs
21.17×
Reported
license unknown · no source
submission 421190Sheheryar Ahmad
python
242.6µs
21.23×
Reported
license unknown · no source
submission 488289yolojewjitsu
python
243.1µs
21.28×
Reported
license unknown · no source
python
243.3µs
21.30×
Reported
license unknown · no source
submission 488072rylanmalarchick
python
243.5µs
21.32×
Reported
license unknown · no source
python
244.4µs
21.39×
Reported
license unknown · no source
python
247.2µs
21.64×
Reported
license unknown · no source
python
275.9µs
24.15×
Reported
license unknown · no source
python
286.5µs
25.08×
Reported
license unknown · no source
python
310.0µs
27.13×
Reported
license unknown · no source
python
320.3µs
28.04×
Reported
license unknown · no source
python
322.5µs
28.23×
Reported
license unknown · no source
python
326.4µs
28.57×
Reported
license unknown · no source
python
326.4µs
28.57×
Reported
license unknown · no source
submission 398933Clark Kitchen
python
331.4µs
29.01×
Reported
license unknown · no source
python
387.2µs
33.89×
Reported
license unknown · no source
python
859.4µs
75.22×
Reported
license unknown · no source
python
901.9µs
78.93×
Reported
license unknown · no source
python
1.23ms
107.65×
Reported
license unknown · no source
submission 488975我爱拆拆
python
1.74ms
152.65×
Reported
license unknown · no source
python
2.07ms
181.03×
Reported
license unknown · no source
python
4.97ms
434.77×
Reported
license unknown · no source
submission 473020kaiming-cheng
python
9.12ms
798.52×
Reported
license unknown · no source
python
9.36ms
818.90×
Reported
license unknown · no source
submission 409296我爱拆拆
python
9.41ms
823.87×
Reported
license unknown · no source
submission 409270我爱拆拆
python
9.46ms
827.95×
Reported
license unknown · no source
submission 383494Clark Kitchen
python
9.54ms
834.77×
Reported
license unknown · no source

Semantics

Inputs and outputs
anvfp4 [m, k, l]
bnvfp4 [n, k, l]
sfafp8_e4m3 [m, k_scales, l]
sfbfp8_e4m3 [n, k_scales, l]
cfp16 [m, n, l]
Axes and behavior
gvariable
kvariable
lconstant = 1
mvariable
nvariable
seedvariable
k_scalesk_scales = k // 16
determinismunspecified
constraintsNo mutation or aliasing
Identity
aliasnvfp4 group gemmaliasnvfp4_group_gemmsha256502b57c10fc1…
Sources: GPU MODE KernelBot (2026-02-21) · June 9 Researcher Reciprocity License v1.0last observed 2026-02-21How records are decidedJSON