MAX Nightly 26.3.0.dev2026032405 Released

:astronaut: A new nightly version has been released! :astronaut:

See the quickstart guide for installation instructions: Quickstart | Modular

MAX changelog updates:

  • Added TaylorSeer denoising cache support to the FLUX.2 Klein pipeline,
    enabling significant speedups for image-to-image generation by skipping
    redundant transformer passes during the denoising loop.

  • Consolidated KV connector CLI flags (--host-kvcache-swap-space-gb,
    --disk-offload-dir, --disk-offload-max-gb, --disk-offload-direct-io,
    --lmcache-config-file) into the --kv-connector-config JSON dict.

  • Removed individual KV connector CLI flags (--host-kvcache-swap-space-gb,
    --disk-offload-dir, --disk-offload-max-gb, --disk-offload-direct-io,
    --lmcache-config-file). Use --kv-connector-config with a JSON dict
    instead.

  • Optimized GPU layer_norm kernels with SIMD reductions, gamma/beta
    prefetch, and a simd_width*2 warp tiling dispatch path.

Mojo changelog updates:

  • [stdlib] Reenable implicit conversion between Span and List
  • [lint] Auto-apply rumdl fixes
  • [mojo-tooling] Elide _mlir_value wrappers and scalar [0] indices in LLDB
  • [mojo-stdlib] Hackaround difficulty in the List->Span ctor.

Raw MAX diff: https://github.com/modular/modular/compare/48ea9af7e1f43e6f6636051a66cf53fd2a87514d...62c6aea99940d0c2d3e70e354924bd9268ba0802)>
Current Mojo changelog: https://github.com/modular/modular/blob/main/mojo/docs/nightly-changelog.md