Key Info

MiniMax responded to H3 being brought to TPUs via SGLang-TorchTPU, a TPU plugin for SGLang built with the Google TPU team and Meta's PyTorch team, which lists MiniMax H3 among top open models supported on TPU.

Highlights

  • SGLang-TorchTPU is a TPU plugin for SGLang developed with Google and Meta PyTorch teams.
  • It is already powering production inference at Meta.
  • Supported top open models include Muse Spark Open, GLM-5.3 Flash, Kimi K3, Ling 3.1 Flash and MiniMax H3.
  • MiniMax highlights benefits for efficient video inference: faster creative iteration and new interactive experiences.