# Google gooses Python AI performance by... removing Python

DevFeed: [Google gooses Python AI performance by... removing Python](<https://devfeed.tech/articles/google-gooses-python-ai-performance-by-removing-python-129517.md>)

Original publisher: [Read original article](<https://blog.adafruit.com/2026/11/11/google-gooses-python-ai-performance-by-removing-python/>)

Author: Anne Barela

Published: 2026-11-11T20:08:22Z

Content type: news

Language: en

Sources: [Adafruit Industries](<https://devfeed.tech/sources/adafruit-industries-makers-hackers-artists-designers-and-engineers.md>)

Topics: [Google](<https://devfeed.tech/topics/google.md>), [Python](<https://devfeed.tech/topics/python.md>), [Model Development](<https://devfeed.tech/topics/model-development.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-performance](<https://devfeed.tech/tags/ai-performance.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [c-plus-plus](<https://devfeed.tech/tags/c-plus-plus.md>), [code](<https://devfeed.tech/tags/code.md>), [engine](<https://devfeed.tech/tags/engine.md>), [google](<https://devfeed.tech/tags/google.md>), [interpreter](<https://devfeed.tech/tags/interpreter.md>), [jax](<https://devfeed.tech/tags/jax.md>), [learning](<https://devfeed.tech/tags/learning.md>), [machine](<https://devfeed.tech/tags/machine.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [math](<https://devfeed.tech/tags/math.md>), [models](<https://devfeed.tech/tags/models.md>), [native](<https://devfeed.tech/tags/native.md>), [open](<https://devfeed.tech/tags/open.md>), [performance](<https://devfeed.tech/tags/performance.md>), [processors](<https://devfeed.tech/tags/processors.md>), [programming](<https://devfeed.tech/tags/programming.md>), [python](<https://devfeed.tech/tags/python.md>), [runtime](<https://devfeed.tech/tags/runtime.md>), [serving](<https://devfeed.tech/tags/serving.md>), [xla](<https://devfeed.tech/tags/xla.md>)

## AI overview

Google's XLA Serving Engine (XSE) serves exported ML models in C++ without Python in the serving path. Models written in Python with JAX are packaged with trained weights as an XLA SavedModel using StableHLO; XLA compiles them into native code for the target chip. Google also plans TorchTPU integration for exporting PyTorch models to C++ serving on TPUs. XSE targets high-rate, latency-sensitive small models for ranking, recommendations, embeddings and vision, rather than large language model inference. Google says it handles over six million queries per second across YouTube Shorts, Google Ads and Search; in its short-request tests, it delivered two to five times the throughput of its best optimized Python server with one quarter the memory. XLA SavedModels can also run on GPUs and CPUs.

## Source excerpt

Joab Jackson at The Register reports that Google has open sourced XSE, the XLA Serving Engine, a C++ runtime for serving machine learning models without a Python interpreter in the path. It's the newest piece of the OpenXLA Project, which compiles machine learning math into native code for different processors. Models are written in Python [...]