inference/research
verifiedhigh · 0.90updated 2026-10-04

Run inference on Cloud TPU

spec-sheet

Fields

id
sources/alt-tpu-inference
Publisher
Google Cloud
Kind
spec-sheet
Published
2026-10-02
Accessed
2026-10-04
notes
Inference on v5e+, single-host up to 8 chips, vLLM tpu-inference plugin for JAX/PyTorch, multi-host via Pathways, serving quotas separate from training.

Notes

Inference on v5e+, single-host up to 8 chips, vLLM tpu-inference plugin for JAX/PyTorch, multi-host via Pathways, serving quotas separate from training.

Cross-references

This record references no other records.