A Reddit developer ported the Radiance image-generation engine to WSL2, Windows’ built-in Linux subsystem, on Windows 11, and released the build scripts and patch. What matters is that the barrier to installing local AI is moving beyond “professional users familiar with Linux” toward ordinary developers, although this is still not turnkey software.
What this is
Radiance natively supports only Linux. The developer used WSL2 to run it on Windows, then relied on scripts to build the application automatically, prepare the AMD ROCm runtime environment, and start the service. It ultimately exposes an OpenAI-compatible API—an interface that allows other software to call models in a standardized way—at 127.0.0.1:8000, making it easy to connect existing tools.
The author reports that, on a Radeon RX 9700 32GB and Ryzen 5 7600 with a 131k context length, the service took roughly eight seconds to start and generated text at approximately 75–80 tokens per second for prose and 145–160 tokens per second for code. These figures come from one user on a single hardware configuration and should not be treated as universal performance guarantees.
Industry view
We believe the value of this project lies not in the fact that “Windows can run Linux software,” but in the emergence of a lighter compatibility layer for local model services. Developers can reuse the OpenAI-compatible interface without rewriting an entire application, an approach with particular appeal for small teams and privacy-sensitive use cases.
The risks are equally clear: this is an unofficial port, and the author explicitly states that the upstream project will not merge WSL support. The patch and runtime environment may break when graphics drivers, ROCm versions, or Radiance are updated. This setup has only been verified on an RX 9700, while the stability of other Radeon 9070-series cards remains to be tested by the community. Treating a community patch as a formal product capability is likely to lead users to overestimate their chances of a successful installation.
Impact on regular people
For enterprise IT: Local deployment still adds a layer of system maintenance compared with cloud-based calls. The WSL2 approach can reduce migration costs, but it cannot replace officially supported containers, drivers, and security policies.
For individual professionals: For users willing to tinker with hardware and the command line, local models can provide lower latency and stronger control over data. This, however, is not a one-click installation experience for ordinary office software.
For the consumer market: Advances like this may drive growth in local AI applications, but consumer-grade graphics cards, memory capacity, and model licensing will still determine who can actually use them. In the short term, they are unlikely to change the habits of most people who rely on cloud services.