Modular's 26.4 release enables MAX models to run natively on Apple silicon GPUs, supporting M1 through M5 chips for LLMs, vision, and diffusion models. While performance is optimized for M5 systems with dedicated Neural Accelerators, users are encouraged to report compatibility issues for older hardware.
Background
Modular is developing the MAX platform to unify AI development using its Mojo programming language, aiming to provide high-performance inference across diverse hardware architectures.
- Source
- Lobsters
- Published
- Jun 28, 2026 at 05:21 PM
- Score
- 7.0 / 10