Models
Last updated: 2026-10-07
The Moorhaven Service runs open-weight AI models built by third parties. This page lists each model we serve, who built it, and its license, as section 6 of our Terms of Service describes. We update it when the models change.
| Moorhaven model name | Underlying model | Developer | License |
|---|---|---|---|
moorhaven-galleon-1 | Qwen3.8-27B-FP8 | Qwen (Alibaba Cloud) | Apache License 2.0 |
moorhaven-skiff-1 | Qwen3.8-27B-FP8 | Qwen (Alibaba Cloud) | Apache License 2.0 |
Both model names are currently served by the same underlying model.
To generate text faster, the model server also uses a small companion model, Qwen3.8-27B-DFlash2 (developer: incoai; license: Apache License 2.0), for speculative decoding. It proposes tokens that the main model then checks, so it does not change what the main model would have produced.
All models run on Moorhaven's own servers in the United States. No prompt or response is sent to any model developer.