Nightly: v26.3

This version is still a work in progress.

Highlights

Documentation

MAX models

Add MXFP4 quantization support for GPT-OSS models (e.g openai/gpt-oss-20b).

MAX framework

Inference server

`max` CLI

Python API

Renamed Float8Config to QuantConfig (and related types/functions) to reflect that the config now covers FP8, NVFP4, and MXFP4 quantization.
Renamed related public Python quantization APIs from Float8* names to Quant* names, including parse_float8_config() to parse_quant_config(), and the public quant modules in max.nn and max.pipelines.lib.

Breaking changes

Mojo API

Custom ops

MAX kernels

Mojo language

For all the updates to the Mojo language, standard library, and tools, including all GPU programming and Layout/LayoutTensor changes, see the Mojo changelog

Highlights​

Documentation​

MAX models​

MAX framework​

Inference server​

max CLI​

Python API​

Breaking changes​

Mojo API​

Custom ops​

MAX kernels​

Mojo language​