Independent AI. Models you can put to the test.
A project by Federico Alessandro, 17.
ComRAD 200B — Boreale is Mindral's first published language model: an experimental Mixture-of-Experts release derived from Z.ai's GLM-5.3-Flash.
The public name is ComRAD 200B; the checkpoint contains 194.93 billion parameters. In conversation, it identifies itself as ComRAD Boreale.
| Model | Distribution | License |
|---|---|---|
| 194.93B parameters · MoE | GGUF Q4_K · 114.33 GB | Apache 2.0 |
Model card & downloads · Files · Discussion
Mindral is an independent AI project focused on developing and evaluating language models, then making the resulting checkpoints available for others to test and build on.
ComRAD's internal methodology and compiler remain private. The published checkpoint includes its conversation template, tested inference settings, integrity checks and licensing notices.
ComRAD 200B is an experimental preview. Its model card describes the tested hardware, evaluation scope and known limitations. The current release has been tested with LM Studio's CUDA runtime; broader compatibility and benchmark coverage are still being evaluated.
If you test it, feedback is welcome in the model discussions. A useful report includes the prompt, the response, the runtime and the settings used.
ComRAD contributions are released under Apache 2.0, which permits commercial use, modification and redistribution under its terms. The original Z.ai MIT license and notices are retained with the model.