AI Briefing#h200
The latest AI and developer news about #h200, with the original source and a short summary.
The latest AI and developer news about #h200, with the original source and a short summary.
The Tang project uses an IR-based compiler to derive GPU kernels, achieving 5-12% faster decode steps for 80B MoE models at high concurrency and reducing 44k-token prefill time by a third on H200 GPUs.
Concurrency tests on H200 GPUs show the model retains its efficiency edge, achieving 985 correct answers per hour at 128 users versus 616 for the base model.
StepFun released the weights for its 600B MoE model Step 5 Preview and confirmed it can run on 8 H200s.
In the GPU rental market, the supply of multiple identical GPU clusters actually needed for real workloads is far more limited than single-GPU supply, and the pricing structure differs as well.
NVIDIA B200 GPU spot prices surged 114% in six weeks.
They transformed manual, legacy AI operations through Kubeflow and GPU optimization.