Halowin Bubo docs

These guides cover how to install, run and tune Halowin Bubo on Windows.

Start here

1. Set up your system.
2. Install Halowin Bubo.
3. Use the app, or read the quickstart for the command line.

Recommended settings

Use Qwen3.8 Flash Next in GGUF format with the UD-Q4_K_XL quant. Use Performance mode. The default profile uses optimal settings and medium reasoning effort.

Reference

• Supported models: the models that were tested on Windows.
• Profiles: settings for each use.
• Options: every server option and environment value.
• API reference: every HTTP endpoint, with request fields, streaming, tool calls and errors.
• Fast restarts with the disk cache: the optional disk prompt cache.
• Known issues: memory, tracing, and the disk prompt cache.
• Benchmarks: measured speeds and test conditions.
• NPU prefill: an option, off by default, that uses the Ryzen AI NPU for part of the prompt processing.
• How it works: the features behind the speed, with the measured numbers.