FULL STORY

Modular Fully Open-Sources Mojo and MAX at ModCon 2026

At the first ModCon after Qualcomm's acquisition, Chris Lattner laid out a unified AI software stack vision and announced that Mojo and MAX are now fully open-sourced under Apache 2.0.

2026-08-19 ~ 2026-08-19 · 2 episodes · 12 posts

Episode 1 · ModCon 2026: Mojo Fully Open-Sourced as Modular Launches Cross-Vendor Alliance (2026-08-19, 8 posts)

At ModCon 2026, Modular and Qualcomm announced that the Mojo language is now fully open source under Apache 2.0 (with LLVM exception), with the MAX platform opened as well and a new open alliance formed to unify AI silicon across vendors. This gives heterogeneous computing a genuinely open software platform and could reduce dependence on closed, single-vendor AI infrastructure stacks.

Confirmed

  • License is Apache 2.0 with an LLVM exception (@DynamicWebPaige)
  • The open-source release covers the Mojo compiler, toolchain and all sources needed to build the language, plus Modular Platform components including the MAX framework, hosted in Modular's official repositories (@DynamicWebPaige, @carrycooldude)
  • Mojo was previously released in partially open form; this marks full open sourcing, with Mojo 1.0 included per @clattnerllvm's announcement
  • Modular Platform already supports NVIDIA, AMD, AWS Trainium, Google TPU and more (@clattnerllvm)
  • Chris Lattner announced opening MAX and forming an open alliance to unify NVIDIA, AMD, Trainium, Google TPU, Qualcomm Dragonfly and Apple Silicon, supporting hardware innovators and researchers (@clattnerllvm)
  • ModCon '26's theme was "heterogeneous computing now has a real software platform"; videos will be published later (@clattnerllvm)

Why it matters

  • Mojo, from Chris Lattner's Modular, targets unified AI development and deployment: Python's usability with C++-level performance, optimized for AI compute bottlenecks and seamless Python migration with large speedups.
  • Full open sourcing (including MAX and multi-hardware backends) plus a cross-vendor alliance could reduce reliance on closed, hardware-locked AI infrastructure software and accelerate the heterogeneous computing ecosystem.

Episode 2 · ModCon: Modular unveils unified AI compute stack with Qualcomm accelerators (2026-08-19, 4 posts)

At ModCon, Modular—now acquired by Qualcomm—positioned its stack as a 'hypervisor for AI compute,' supporting any model on any accelerator. Qualcomm's data center AI accelerator roadmap was integrated, letting one codebase run across Nvidia, AMD, and NPUs, while CEO Cristiano Amon shared his AI inference ecosystem vision.