NInfer 5090 Windows
Native Windows port of the NInfer C++/CUDA inference engine, built for a single RTX 5090 GPU with no WSL2 or Docker, serving text, image and video prompts through a local CLI or OpenAI/Anthropic-compatible HTTP APIs.
NInfer 5090 Windows is a native Windows build of the NInfer engine, compiled with MSVC and CUDA 13.3 rather than through WSL2 or Docker. It targets a single RTX 5090 GPU with one resident model and a fixed capacity of one to eight active requests.
It serves text, image and video prompts through a local CLI or through HTTP APIs compatible with the OpenAI and Anthropic formats. Release v1.1.0 ships two vision-capable Qwen3.8-27B artifacts, a QUASAR QAT build and an NVFP4-full build, both supporting a 262,144-token context with DFlash2 or MTP speculative decoding.
Installation downloads a release archive checked against a SHA-256 hash; a bundled script fetches and verifies the model artifact, and a launcher starts a local server exposing a /v1 API and a /health endpoint. The engine requires v3 artifacts and rejects older v2 ones.