Qwen3模型在GB200平台实现高效推理

Perplexity@perplexity_ai

2026-05-12 22:17·34天前

AI 摘要

我们发布了关于如何在NVIDIA GB200 NVL72 Blackwell机架上部署训练后Qwen3 235B模型的新研究。 GB200不仅是训练平台，更为大型MoE模型的高吞吐量推理带来了重大升级，相比Hopper实现显著进步。

We published new research on how we serve post-trained Qwen3 235B models on NVIDIA GB200 NVL72 Blackwell racks.

GB200 is a major step up over Hopper for high-throughput inference on large MoE models， not just a training platform.

论文/研究部署/工程

Perplexity@perplexity_ai · X

2026-05-12 22:17·34天前

AI 摘要

We published new research on how we serve post-trained Qwen3 235B models on NVIDIA GB200 NVL72 Blackwell racks.

GB200 is a major step up over Hopper for high-throughput inference on large MoE models， not just a training platform.

论文/研究部署/工程