--- title: README emoji: ⬛ colorFrom: gray colorTo: gray sdk: static pinned: false ---

Sectile Research Laboratories. We build the model, and everything it runs on.

**Sectile Research Laboratories** is rebuilding the AI stack from the representation up. We release consumer-focused versions of the systems we run on to the open-source and local-hosting community.

GInfer: open-source inference engine for fast, high-concurrency, long-context work on NVIDIA GPUs, Apache 2.0. GChat: the desktop app for GInfer on Windows and Linux.

## Models The [GInfer Models](https://huggingface.co/collections/SectileLabs/ginfer-models-6aa0682395e628be98a798ef) collection holds our model packages for GInfer and GChat. Each package's card names its base model and license. | Format | What it is | GPUs | |---|---|---| | INT | 4-bit groupwise integer | Ampere, Ada, Blackwell | | NVFP4 | 4-bit NVIDIA FP4 | Blackwell (RTX 50 series, RTX PRO) | | SMOL | 2-bit, mixed with 4-bit where a model needs it | Ampere, Ada, Blackwell | The packages run only in GInfer and GChat. ## Learn more - [sectilelabs.ai](https://sectilelabs.ai) - [Case study: building GInfer with Quaestor](https://sectilelabs.ai/case-study-ginfer.html)