Razer AIKit expands with new models and other improvements

Razer has expanded its open‑source AI development toolkit, Razer AIKit, adding native support for image, video, and audio models while broadening compatibility to include Arm64 systems. First introduced at CES 2026, the update shifts AIKit from a text‑focused toolkit into a fully omni‑modal development platform designed to run advanced models locally across a wide range of hardware, including recent Blade laptops.

Razer says the goal is to streamline development as projects scale, with AIKit handling GPU discovery, configuration, and performance tuning automatically. The company also highlighted AIKit’s role in powering AVA Mini, its April Fool’s campaign that let users generate personalised 3D AI pet companions in seconds. The same toolkit used during development supported the live deployment.

The update adds support for image‑generation models such as Tongyi‑MAI/Z‑Image‑Turbo and FLUX.2‑klein‑base‑4B, enabling developers to prototype and deploy generative features through a unified workflow. Arm64 support extends AIKit to systems built on NVIDIA’s latest professional platforms, including DGX Spark and hardware powered by Grace Hopper and Grace Blackwell Superchips, allowing teams to operate consistently across x86 and Arm environments.

For AVA Mini, Razer partnered with Akash Network to run AIKit across a distributed pool of consumer GPUs, reducing inference costs to US$0.01 per image. The five‑day campaign generated over 11,000 images with an average turnaround of 3.24 seconds and required no manual intervention.

A joint whitepaper from Razer and Akash Network going over all of this in more detail is available to read, HERE.

KitGuru Says:

Become a Patron!