List your company

LLM Training Performance on Qwen models on A6000 (48GB VRAM) 2.5 (train)

56.82TFLOPSMegaTrain
22.572831.463940.35549.2461Apr 6, 2026

Evaluation results

MethodLinks
MegaTrainModel=Qwen2.5-14B, Max BS=92026.0456.8244.64104.1641
MegaTrainModel=Qwen2.5-7B, Max BS=122026.0455.7344.7456.71,219
MegaTrainModel=Qwen2.5-3B, Max BS=152026.0449.746.7438.32,153
ZeRO-3 OffloadModel=Qwen2.5-7B, Max BS=12026.0427.5520.83——
ZeRO-3 OffloadModel=Qwen2.5-3B, Max BS=12026.0423.8920.33——