LynxBenchAI Training Leaderboard

AI hardware re-ranked by training-category score — the throughput profile that matters when the workload is model training rather than inference or raw compute. GT and the other two category scores stay visible for context. Results: 26Q3.

# Device Accelerator / GPU GT Global Score Training Category Inference Category Compute Category Platform OS
1
1306.35
1089.67
2224.55
611.10
Native Linux
2
1277.42
1080.48
2105.93
624.58
Native Linux
3
1044.99
811.98
1914.61
480.83
Native Linux
4
986.72
727.23
1304.04
1023.28
Native Linux
5
939.45
709.35
1678.11
487.13
Native Linux
6
692.67
588.50
1123.94
347.75
Native Linux
7
718.86
587.80
1201.04
366.56
Native Linux
8
597.49
515.26
985.84
279.64
Native Linux
9
440.29
415.55
660.97
210.04
Native Linux
10
403.38
387.30
584.07
201.12
Native Linux
11
450.05
335.94
659.37
367.89
Native Linux
12
395.82
316.04
528.12
344.33
Native Linux
13
202.73
207.55
294.13
87.37
Native Linux
14
160.52
163.70
240.49
64.77
Native Linux
15
158.79
157.40
231.75
72.52
WSL
16
146.10
133.08
239.43
61.68
Native Linux
17
74.55
60.71
100.85
60.59
WSL
18
76.32
37.28
122.73
112.62
Native Linux
19
5.20
8.58
2.69
6.32
Native Linux
20
6.35
8.52
5.78
4.19
WSL
21
7.87
6.69
9.05
8.30
Native Linux
22
0.34
0.70
0.34
0.07
WSL
23
0.30
0.20
0.17
1.59
WSL

The Personal Edition submits your benchmark results to TechnoLynx's servers as part of participating in the public leaderboard. If that isn't acceptable for your use case, contact TechnoLynx about Press, Pro, or Enterprise licenses.

Get Started

Results: 26Q3 · ranked by training-category score, higher is better. GT (Global Score) is LynxBenchAI's overall crown; this view answers "which hardware leads specifically on training," not "which hardware wins overall." Bars are scaled to each column's own leader, not a shared ceiling -- compare bars down a column, not across.