Modular 26.2: State-of-the-Art Image Generation and Upgraded AI Coding with Mojo

Modular Official Announcements
1

26.2 is live! Here are the can’t-miss updates.

The big one for kernel developers: Mojo :fire: now has AI coding skills specifically for writing high-performance, portable GPU kernels. Combined with a growing base of 750K+ lines of open-source Mojo kernel code, agents have real context to work from. Shorter feedback loops, compile-time errors, readable output.

Also in the release: FLUX.2 image generation lands in MAX (3-4x faster than PyTorch Diffusers, no infra changes needed if you’re already on MAX), plus AMD RDNA and NVIDIA B300 support and Qwen3-30B out of the box.

Check out the full blog post: Modular: Modular 26.2: State-of-the-Art Image Generation and Upgraded AI Coding with Mojo