| license: apache-2.0 | |
| base_model: Qwen/Qwen3.8-27B | |
| library_name: basert | |
| pipeline_tag: text-generation | |
| tags: | |
| - basert | |
| - apple-silicon | |
| - quantized | |
| # Qwen3.8-27B | |
| [BaseRT](https://github.com/basecompute/baseRT) `.base` builds of [`Qwen/Qwen3.8-27B`](https://huggingface.co/Qwen/Qwen3.8-27B) for fast local inference on Apple Silicon (Metal). | |
| A dense 27B vision-language reasoning model with hybrid attention (Gated DeltaNet + periodic full attention). Q4 needs ~24 GB Apple Silicon (tight); Q8 needs 48 GB+. | |
| ## Files | |
| | File | Precision | Size | | |
| |------|-----------|------| | |
| | `Qwen3.8-27B-Q4.base` | 4-bit | 17 GB | | |
| | `Qwen3.8-27B-Q8.base` | 8-bit | 28 GB | | |
| ## Usage | |
| ```bash | |
| curl -LsSf https://basecompute.co/install.sh | sh | |
| basert pull basecompute/Qwen3.8-27B | |
| basert chat basecompute/Qwen3.8-27B | |
| ``` | |
| Released under the apache-2.0 license, inherited from the base model. | |