What Is Razer AIKit?

Razer's open-source local AI toolkit — what it is, what it isn't, and how it relates to Razer AVA.

A local AI toolkit, not an AVA-exclusive SDK

Razer AIKit is an open-source platform Razer announced at CES 2026 (January 6, 2026) that simplifies local AI development: it auto-discovers and configures GPUs, forms clusters, and optimizes inference and fine-tuning for language models without relying on the cloud. It is a general-purpose tool for AI developers and researchers — it was not announced as "the Project AVA SDK", though it shares the same maker and ecosystem.

Official repository

Official source code and documentation: github.com/razerofficial/aikit

What does it actually do?

AIKit combines three well-known pieces of the open-source AI stack: vLLM for fast inference, Ray for distributed computing and multi-GPU scaling, and LlamaFactory for efficient fine-tuning. It gives access to 300,000+ models from the Hugging Face Hub, and since its GDC 2026 expansion (May 2026) it also supports omni-modal models (image, video, audio) plus ARM64 architectures.

Local inference

Run language models directly on your GPU without sending data to cloud services.

Multi-GPU via Ray

Automatically orchestrates GPU clusters to scale processing without manual setup.

Omni-modal (since GDC 2026)

Beyond text: supports image, video and audio models in a single workflow.

Open source

Published under the Apache 2.0 license on GitHub — inspect, modify and contribute freely.

Is it the Razer AVA SDK?

Not exactly, and precision matters here: Razer has not officially stated that AIKit is "the Project AVA SDK". They are two related but distinct Razer AI initiatives. AVA (the desktop hologram companion) currently runs on xAI's Grok as its AI backend, though Razer has said its architecture is open to integrating "a dedicated Razer AI" in the future — which could plausibly build on AIKit, but that is not confirmed.

What is confirmed: AIKit powered the image-generation workload for the "AVA Mini" campaign (30 images/minute across distributed GPUs via Akash Network, at $0.01/image) — a specific collaboration, not a declared permanent architecture.
AVA SDK (avasdk.com) is an independent community resource: we track and document Razer's local AI ecosystem for developers interested in Project AVA, without being the official channel for either product.
Independence

AvaSDK.com is an independent, community-run educational resource. We are not affiliated with, endorsed by, or officially connected to Razer Inc.

Verified requirements

NVIDIA GPU with Compute Capability 7.0 or higher (Volta, Turing, Ampere, Ada Lovelace, Hopper or Blackwell)
Windows 11 (via WSL2) or Ubuntu 22.04 / 24.04
Docker Engine + NVIDIA Container Toolkit
Up-to-date NVIDIA GPU driver (Razer recommends the Studio Driver on Windows)
No Razer hardware required — runs on any compatible GPU system, though it is optimized and tested on Razer Blade laptops

Getting started (official Docker Quick Start)

1

mkdir -p $HOME/.cache/huggingface

2

docker run -it --restart=unless-stopped --gpus all --ipc host --network host --mount type=bind,source=$HOME/.cache/huggingface,target=/var/aikit/.cache/huggingface --env HUGGING_FACE_HUB_TOKEN=<YOUR_TOKEN> razerofficial/aikit:latest

3

Inside the container: launch Jupyter Lab or run models directly via CLI

4

For the advanced Docker Compose setup: git clone https://github.com/razerofficial/aikit.git && cd aikit && docker compose -f docker_compose/docker-compose.yaml up -d --pull always

AIKit on Linux (Ubuntu 22.04 / 24.04)

AIKit runs natively on Ubuntu, without the WSL2 layer Windows needs — setup is more direct because you skip the networking/firewall steps WSL2 requires for GPU access on Windows.

NVIDIA driver: install it via Software & Updates → Additional Drivers (no need for the "Studio Driver" recommended on Windows)
Verify the GPU is detected correctly with nvidia-smi before continuing
Install Docker Engine (official Docker guide) and the NVIDIA Container Toolkit so containers can access the GPU
From there, the same docker run command from the Quick Start works exactly as it does on Windows/WSL2

Want more guides like this?

Subscribe and we'll let you know as soon as Razer shares news about AVA and its local AI ecosystem.

Join

Frequently asked questions

Is Razer AIKit free?

Yes. It is open-source under the Apache License 2.0, available at github.com/razerofficial/aikit at no cost.

Is it the same as Project AVA / Razer AVA?

No. Project AVA is the holographic desktop companion; its current AI engine is xAI's Grok. AIKit is Razer's general local-AI development toolkit. They are sibling products from the same maker, not the same thing.

Do I need a Razer device to use it?

No. It runs on any PC with a compatible NVIDIA GPU (Compute Capability 7.0+). It is optimized for Razer Blade laptops, but that is not a requirement.

Is there an official "Razer SDK" for modding AVA?

No, there is no public modding SDK for Project AVA as of today. The closest things are AIKit (for local AI in general) and Razer's character/IP integration program for studios — see the details on our Razer AVA SDK page.

Sources