Wayland 0.11.4 ships Crucible, a Mixture-of-Providers council that convenes multiple models from different providers on one task, lets each take a swing, and fuses the strongest answer. Flux Auto routes every request to the best model for you, chat streaming is greatly improved, and real tool-use runs sandboxed on Windows.
The bundled engine now packs Crucible, a cross-provider council that convenes multiple models on one task, lets each take a swing, and fuses the strongest answer. Flux Auto routes each request to the best model automatically, so you stop guessing which AI to use.
An inline orbit indicator animates while the assistant runs, a unified plain-English activity timeline reads the same across every backend, and live reasoning summaries land in the thinking block.
Wayland runs real shell commands and tools on Windows inside a locked-down AppContainer sandbox, working on a clean non-developer machine. Live-verified.
Ollama and llama.cpp models run private and free, and now handle tools, thanks to up-front capability detection with a graceful no-tools fallback.
Every connected provider shows an accurate, live model catalog, so the picker always reflects the current lineup for each backend you run.
Assistants run on the Wayland Core engine by default, so the full toolset, sandbox, and routing are there from the first message.
Grok "Sign in with X" stays authenticated through a single race-free OAuth refresher, so your xAI session holds without re-prompting.
One-click assistants land with starter-prompt grids, multi-model teams coordinate the work, and scheduled tasks keep recurring runs moving.
Already running Wayland? Update from the pill in the title bar. New here? Grab the latest build, free and open source.