Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
LLM Training Performance on Qwen models on RTX 3090 (24GB VRAM) 2.5 (train)
Loading...
35.09
TFLOPS
MegaTrain
23.4628
26.4814
29.5
32.5186
Apr 6, 2026
TFLOPS
GPU Memory (GB)
CPU Memory (GB)
Throughput (tokens/s)
Updated 1mo ago
Evaluation Results
Method
Method
Links
TFLOPS
GPU Memory (GB)
CPU Memory (GB)
Throughput (tokens/s)
MegaTrain
Model=Qwen2.5-7B, Max...
2026.04
35.09
22.63
56.7
768
MegaTrain
Model=Qwen2.5-3B, Max...
2026.04
33.18
22.83
25
1,792
MegaTrain
Model=Qwen2.5-14B, Max...
2026.04
30.19
21.1
103.7
341
ZeRO-3 Offload
Model=Qwen2.5-7B, Max...
2026.04
27.49
20.83
-
-
ZeRO-3 Offload
Model=Qwen2.5-3B, Max...
2026.04
23.91
20.32
-
-
Feedback
Search any
task
Search any
task