Editions that work for everyone.
‍Develop and scale with us.

MAX is a high performance inference framework that enables you to serve LLMs, write high performance inference graphs, and develop custom, scalable test-time inference pipelines across image, audio, video and text.

‍

Mojo is a pythonic language for blazing-fast CPU+GPU execution without CUDA. Optionally use it with MAX for insanely fast AI inference.

You can use Mojo and MAX for any non-production or non-commercial use, however you’d like.

‍

Additionally, you can use Mojo and MAX for free for commercial purposes - we ask that you let us know and allow us to put your company logo on our website so we can demonstrate the impact our work has by emailing usage@modular.com. Such usage is unlimited on CPUs and NVIDIA hardware, and up to 8 discrete GPUs from other vendors - though MAX doesn’t support any other vendors today.

Yes, for any purpose.

‍

If you deploy MAX & Mojo for commericial, production use - we ask that you let us know and allow us to put your company logo on our website so we can demonstrate the impact our work has by emailing usage@modular.com. Such usage is unlimited on CPUs and NVIDIA hardware, and on up to 8 discrete GPUs from other vendors - though MAX doesn’t support any other vendors today.

Yes, most laptops and desktops have CPUs and 8 or fewer discrete GPUs in them - and you can use MAX for anything on such machines. Additionally, you can use Mojo and MAX for free for any non-production or non-commercial use-case!

‍

Please just Talk to Us if you want to use MAX for more than 8 non-NVIDIA accelerators in a PC for some commercial purpose.

Yes, Apple GPUs show up as a single device in the computer, and therefore are counted as a single GPU, so you can use them unlimited for both personal and commercial use-cases.

‍

Keep in mind that MAX does not support Apple GPUs today, but this is something we may explore in the future.

Modular software is licensed on a discrete physical per-device basis. So any discrete device accessible by the software is what we consider as "a single device".

‍

The Community Edition enables you to scale limitlessly for non-production, non-commercial usage cases. For production, commercial use you can scale to unlimited CPU, NVIDIA hardware deployments, and can utilize up to 8x other device types before we require you to contact us for licensing.

The Modular Community License is intentionally structured so Modular can’t “rug pull” the software.

‍

If for some reason Modular decides to stop updating the SDK, you can continue use of the existing versions and get a perpetual license to do so. This is much more liberal than licenses for comparable software like CUDA.

We would like to be able to scale to support a wide variety of hardware from multiple vendors over time, but AI software is difficult to build and expensive for a startup like us to scale, support, and qualify in our releases. 

‍

MAX is free for use on all hardware it officially supports today, and we will continue to evaluate the right approach on a case-by-case basis over time as we expand support for more hardware.

‍

If you’re interested in MAX supporting hardware from other vendors, please reach out to them and encourage them to work with us.

‍